Arena.ai(@lmarena_ai)

We analyzed how similar model responses were across 30,086 Arena battles. Models shared 43% of thei...

Arena.ai (@arena) — Fri, 11 Sep 2026 18:23:03 GMT

We analyzed how similar model responses were across 30,086 Arena battles.

Models shared 43% of their ideas on average. We might expect that models from the same lab, or country, would show greater conceptual overlap. But the results don’t consistently support that.

Claude Fable 5 illustrates this pattern: its closest conceptual match was neither Opus nor Sonnet. Which model came closest, along with the broader findings, may surprise you.

Check out the full article from @DawidGalarowicz and @petergostev below.

Arena.ai@arena

x.com/i/article/2098…

🔗 View Quoted Tweet

💬21🔄20❤️201👀21078📊39

⚡ Powered by xgo.ing

https://t.co/K3TE6Bj6pZ

Arena.ai (@arena) — Fri, 11 Sep 2026 18:01:23 GMT

x.com/i/article/2098…

💬6🔄5❤️106👀37297📊28

⚡ Powered by xgo.ing

Muse Spark 1.3 (Max) increased performance across all key signals vs. Muse Spark 1.2 (xHigh): - Conf...

Arena.ai (@arena) — Fri, 11 Sep 2026 14:21:20 GMT

Muse Spark 1.3 (Max) increased performance across all key signals vs. Muse Spark 1.2 (xHigh):
- Confirmed Success: +8.4% (vs. +1.5%)
- Praise vs. Complaint: +5.3% (vs. -8.4%)
- Steerability: +1.4% (vs. -7.9%)
- Bash Recovery remained at +5.7%
- No issues with Tool Hallucination

💬1🔄0❤️6👀3578📊2

⚡ Powered by xgo.ing

Agent Arena measures models on millions of real-world, long-horizon agentic tasks. Models use web se...

Arena.ai (@arena) — Fri, 11 Sep 2026 14:21:20 GMT

Agent Arena measures models on millions of real-world, long-horizon agentic tasks. Models use web search, filesystem, and terminal tools to complete complex workflows.

Using causal tracing, net improvement indicates how much a model improves outcomes relative to the average model.

See the full Agent Arena leaderboard: arena.ai/leaderboard/ag…

💬0🔄0❤️7👀2909📊1

⚡ Powered by xgo.ing

Muse Spark 1.3 (Max) by @AIatMeta lands at #13 in Agent Arena with +4.2% net improvement across 8.7K...

Arena.ai (@arena) — Fri, 11 Sep 2026 14:21:19 GMT

Muse Spark 1.3 (Max) by @AIatMeta lands at #13 in Agent Arena with +4.2% net improvement across 8.7K+ real-world agentic sessions.

That’s a 17-rank climb from Muse Spark 1.2 (xHigh) at #30, and a +6 percentage-point lift from −1.8%.

Muse Spark 1.3 (Max) also improved from Muse Spark 1.2 (xHigh) across every category:
- Code: #15 (+4.3%), up from #29 (-2.5%)
- Work: #15 (+3.7%), up from #30 (-3.1%)
- Chat: #17 (+1.6%), up from #17 (+1.2%)
That momentum carries into its signal-level results. See the full breakdown below.

Congrats to the @AIatMeta team on this release!

AI at Meta@AIatMeta

We’re excited to release Muse Spark 1.3 with improved performance on agentic and coding tasks, and a focus on real-world usability.

Key capabilities:
→ Sustains longer-horizon work across multiple workflows in a single thread
→ More actively collaborates with users: it asks clarifying questions, flags when it's stuck, confirms before consequential actions
→ Better calibrated on its own limits instead of hallucinating outcomes
→ ~20% fewer tool calls and ~25% fewer tokens vs. Muse Spark 1.2 in internal comparisons

🔗 View Quoted Tweet

💬19🔄12❤️250👀28134📊36

⚡ Powered by xgo.ing

Use GPT-Image-2.5 Sunburst in Direct Mode on Arena now: https://t.co/tYRvUy9yTd https://t.co/rzljW...

Arena.ai (@arena) — Fri, 11 Sep 2026 03:02:41 GMT

Use GPT-Image-2.5 Sunburst in Direct Mode on Arena now: bit.ly/4xNVAA8x.com/arena/status/2…

Arena.ai@arena

For the next 72 hours on Arena, you can use GPT-Image-2.5 Sunburst directly at no cost.

@OpenAI’s latest image model was just  ranked #1 in the Text-to-Image Arena, the Image Edit Arena, and in the Multi-Image Edit Arena. Find the link below, select the model in Direct Mode, and test it directly for yourself!

Direct Mode lets users directly select models by name. Access to GPT-Image-2.5 Sunburst ends Sept. 13 at 8 a.m. PT. Afterward, it will remain available anonymously in Battle and Agent Mode. User votes from Battle and Agent Mode are used to power our real-world leaderboard rankings. In these modes, models are always anonymized.

🔗 View Quoted Tweet

💬1🔄0❤️15👀7505📊2

⚡ Powered by xgo.ing

In case you missed it, GPT-Image-2.5 Sunburst by @OpenAI is #1 in the Image Arena and currently avai...

Arena.ai (@arena) — Fri, 11 Sep 2026 03:02:39 GMT

In case you missed it, GPT-Image-2.5 Sunburst by @OpenAI is #1 in the Image Arena and currently available to use directly for a limited time. Check out how it compares to another top model below, and test for yourself in Direct Mode at the link in the thread. 

Direct Mode access to GPT-Image-2.5 Sunburst ends Sept. 13 at 8 a.m. PT. The model will remain available in Battle and Agent Mode afterward.

youtube.com/shorts/Y-0K-Vz…

💬11🔄4❤️100👀18367📊21

⚡ Powered by xgo.ing

Our CEO and Co-Founder, @ml_angelopoulos, joined @kleinerperkins for a new episode of Grit. They di...

Arena.ai (@arena) — Thu, 10 Sep 2026 20:47:37 GMT

Our CEO and Co-Founder, @ml_angelopoulos, joined @kleinerperkins for a new episode of Grit.

They discuss @arena's real-world vantage point on the rapid pace of AI progress: including how open models are closing key performance gaps, Chinese labs are advancing the frontier, and more.

youtube.com/watch?v=bwxhra…

💬5🔄1❤️44👀29105📊10

⚡ Powered by xgo.ing

Agent Arena measures models on millions of real-world, long-horizon agentic tasks. Models get web se...

Arena.ai (@arena) — Thu, 10 Sep 2026 18:30:15 GMT

Agent Arena measures models on millions of real-world, long-horizon agentic tasks. Models get web search, filesystem, and terminal tools to complete complex workflows: writing code, creating slide decks, researching the web, building apps, and analyzing documents.

We use causal tracing methodology to measure a model's net improvement, which indicates how much it improves outcomes relative to the average model.

Dive into Agent Arena at: arena.ai/leaderboard/ag… and check out the Pareto frontier: arena.ai/leaderboard/ag…

💬1🔄0❤️6👀4036📊2

⚡ Powered by xgo.ing

Last month’s Hy4 preview launch landed @TencentHunyuan the #2 spot on the top 10 labs in Agent Arena...

Arena.ai (@arena) — Thu, 10 Sep 2026 18:30:14 GMT

Last month’s Hy4 preview launch landed @TencentHunyuan the #2 spot on the top 10 labs in Agent Arena among open-source labs!

Across 14.5K+ real-world agent sessions, Hy4 preview ranks #2 among open models (#10 overall) with a +5.4% net improvement, trailing only Kimi K3 (Max) at +6.6%—a gap of just nearly 1 percentage point.

This level of performance comes at roughly 68% lower cost: $0.26 per median task, compared with $0.80 for Kimi K3 (Max). Hy4 preview is now on the Agent Arena Pareto frontier. See the link below to explore its placement.

By signal, Hy4 preview excels in:
- Confirmed Success: 12% — explicit user feedback that the task worked
- Praise vs. Complaint: 9% — implicit sentiment in user reactions
- Bash Recovery: 7.5% — recovery from CLI errors
It shows no issues with Tool Hallucination — calling tools that don’t exist

By category among open models, it’s also #2 in Code and Work, and #3 in Chat.

With this strong performance at a competitive price, Hy4 preview is a huge contribution to the open-source ecosystem. 

Congrats to the @TencentHunyuan team on this strong release!

Tencent Hy@TencentHunyuan

🚀 Hy4 preview is here.

770B, 49B active, 1M context.
Built for productivity.
Open source frontier.
Consistent affordable price.
Use it. Tell us what breaks.

More on Hy bloghy.tencent.ai/research/hy4-p…C
HuggingFacehuggingface.co/tencent/Hy4-pr…R
Githubgithub.com/Tencent-Hunyua…L

🔗 View Quoted Tweet

💬13🔄14❤️181👀23775📊27

⚡ Powered by xgo.ing

See the Code Arena leaderboard with AutoEval score at: https://t.co/GFZ3FCC7Cl and learn more about...

Arena.ai (@arena) — Thu, 10 Sep 2026 16:39:32 GMT

See the Code Arena leaderboard with AutoEval score at: arena.ai/leaderboard/co…  and learn more about AutoEval below

x.com/arena/status/2…

Arena.ai@arena

x.com/i/article/2085…

🔗 View Quoted Tweet

💬0🔄1❤️10👀9765📊1

⚡ Powered by xgo.ing

DeepSeek-V4.1-Flash by @deepseek_ai just landed ~#14 overall in Code Arena: WebDev with 1620 pts (Au...

Arena.ai (@arena) — Thu, 10 Sep 2026 16:39:31 GMT

DeepSeek-V4.1-Flash by @deepseek_ai just landed ~#14 overall in Code Arena: WebDev with 1620 pts (AutoEval)! Among open models, DeepSeek-V4.1-Flash landed at ~#4 within 11 pts of Qwen3.8-Flash-Next.

This release is a significant improvement compared to DeepSeek-V4 variants:
+38 pts vs. V4-Flash (High) at #20 (1582 pts)
+40 pts vs. V4-Pro (High) #21 (1580 pts)

DeepSeek-V4.1-Flash is within 5 pts of the models ranked #11–#13, while costing substantially less (priced per million input/output tokens):
~#14 DeepSeek-V4.1-Flash: $0.30/ $1.20
#13 Hy4 Preview: $0.83/ $2.50
#12 Grok-4.6 High: $2/ $6
#11 Muse Spark 1.3 xHigh: $1.25/ $4.25

Note: this is an early AutoEval score, in which a Reward Model trained on Arena’s human preference data casts automatic votes in place of live votes. We’ll continue to see how scores converge as more live human votes come in.

Congrats to the @deepseek_ai team on this contribution to the open source ecosystem!

DeepSeek@deepseek_ai

🚀 Introducing DeepSeek-V4.1-Flash: smarter, faster, more efficient.

🔹 Introducing the smallest model in our new architecture family, with native visual understanding.
🔹 Designed for greater capability, faster inference, higher throughput, and scaling to larger models.

1/6

🔗 View Quoted Tweet

💬42🔄77❤️1059👀141043📊133

⚡ Powered by xgo.ing

Dive into the Code Arena: WebDev leaderboard details at https://t.co/GFZ3FCC7Cl

Arena.ai (@arena) — Thu, 10 Sep 2026 15:43:05 GMT

Dive into the Code Arena: WebDev leaderboard details at arena.ai/leaderboard/co…

💬0🔄1❤️9👀3861📊1

⚡ Powered by xgo.ing

The frontier gap in web development between proprietary and open source models nearly closed this qu...

Arena.ai (@arena) — Thu, 10 Sep 2026 15:43:04 GMT

The frontier gap in web development between proprietary and open source models nearly closed this quarter, until big releases from @AnthropicAI and @OpenAI last week causing it to snap back open.

From a peak of +150 points in late 2025, the proprietary lead compressed to just approx. +10 points by early Q3 2026 as open source models made large gains.

With Claude Fable 5.1 and GPT-6 Astra, the proprietary frontier jumped back to nearly 1,800 pts while open source held near 1675 pts widening the gap to +120 points.

Votes for DeepSeek-V4.1-Flash by @deepseek_ai are still coming in. Stay tuned for an updated analysis once it lands in the Code Arena: WebDev.

Insights are powered by real-world tasks from users in the Code Arena: WebDev. Link to see for yourself below.

Arena.ai@arena

Code Arena: Frontend is a tighter race.

The history is shorter, and the gap has moved faster: the proprietary lead rose +100 points, then compressed sharply through spring 2026.

Today, the lead is around +40 points.

🔗 View Quoted Tweet

💬10🔄11❤️187👀21211📊25

⚡ Powered by xgo.ing

Head over to find it in Agent Mode and Battle Mode to test it out: https://t.co/Res3BoLgQ3

Arena.ai (@arena) — Thu, 10 Sep 2026 15:32:04 GMT

Head over to find it in Agent Mode and Battle Mode to test it out: arena.ai

💬0🔄1❤️14👀4233📊2

⚡ Powered by xgo.ing

DeepSeek-V4.1-Flash by @deepseek_ai is in the Arena! Bring your toughest prompts and start voting. ...

Arena.ai (@arena) — Thu, 10 Sep 2026 15:32:03 GMT

DeepSeek-V4.1-Flash by @deepseek_ai is in the Arena!

Bring your toughest prompts and start voting. Scores coming soon.

In Agent Arena, we measure models on millions of real-world, long-horizon agentic tasks. Models can access web search, filesystem, and terminal tools to complete complex workflows. The leaderboard measures model performance on outcomes relative to the average model using a causal tracing methodology.

Also available in: the Code Arena: WebDev, the Text Arena, and the Vision Arena.

DeepSeek@deepseek_ai

🚀 Introducing DeepSeek-V4.1-Flash: smarter, faster, more efficient.

🔹 Introducing the smallest model in our new architecture family, with native visual understanding.
🔹 Designed for greater capability, faster inference, higher throughput, and scaling to larger models.

1/6

🔗 View Quoted Tweet

💬13🔄11❤️235👀22053📊28

⚡ Powered by xgo.ing

Use GPT-Image-2.5 Sunburst in Direct Mode on Arena now: https://t.co/OubyEtPPbb

Arena.ai (@arena) — Thu, 10 Sep 2026 15:00:13 GMT

Use GPT-Image-2.5 Sunburst in Direct Mode on Arena now: bit.ly/4hhVKK6

💬1🔄1❤️15👀4288📊4

⚡ Powered by xgo.ing

For the next 72 hours on Arena, you can use GPT-Image-2.5 Sunburst directly at no cost. @OpenAI’s l...

Arena.ai (@arena) — Thu, 10 Sep 2026 15:00:12 GMT

For the next 72 hours on Arena, you can use GPT-Image-2.5 Sunburst directly at no cost.

@OpenAI’s latest image model was just  ranked #1 in the Text-to-Image Arena, the Image Edit Arena, and in the Multi-Image Edit Arena. Find the link below, select the model in Direct Mode, and test it directly for yourself!

Direct Mode lets users directly select models by name. Access to GPT-Image-2.5 Sunburst ends Sept. 13 at 8 a.m. PT. Afterward, it will remain available anonymously in Battle and Agent Mode. User votes from Battle and Agent Mode are used to power our real-world leaderboard rankings. In these modes, models are always anonymized.

OpenAI@OpenAI

ChatGPT Images 2.5—faster, sharper, smarter, with better tools for creating whatever you can dream of.

- Faster image generation to keep your ideas flowing

- Improved fidelity for more natural, recognizable images

- Consistent details across multiple edits

- Comment-based edits to change only what you want

Your browser does not support the video tag.

🔗 View on Twitter

🔗 View Quoted Tweet

💬29🔄34❤️416👀49581📊76

⚡ Powered by xgo.ing

Head over to find it in Agent Mode and Battle Mode to test it out: https://t.co/Res3BoLgQ3

Arena.ai (@arena) — Thu, 10 Sep 2026 14:36:14 GMT

Head over to find it in Agent Mode and Battle Mode to test it out: arena.ai

💬1🔄1❤️9👀3349📊2

⚡ Powered by xgo.ing

DeepSeek-4.1-Flash by @deepseek_ai is in the Arena! Bring your toughest prompts and start voting. S...

Arena.ai (@arena) — Thu, 10 Sep 2026 14:36:13 GMT

DeepSeek-4.1-Flash by @deepseek_ai is in the Arena!

Bring your toughest prompts and start voting. Scores coming soon.

In Agent Arena, we measure models on millions of real-world, long-horizon agentic tasks. Models can access web search, filesystem, and terminal tools to complete complex workflows. The leaderboard measures model performance on outcomes relative to the average model using a causal tracing methodology.

Also available in: Code Arena: WebDev, Text Arena, and Vision Arena.

DeepSeek@deepseek_ai

🚀 Introducing DeepSeek-V4.1-Flash: smarter, faster, more efficient.

🔹 Introducing the smallest model in our new architecture family, with native visual understanding.
🔹 Designed for greater capability, faster inference, higher throughput, and scaling to larger models.

1/6

🔗 View Quoted Tweet

💬9🔄13❤️181👀12031📊23

⚡ Powered by xgo.ing

We analyzed how @claudeai's writing has changed from Fable 5 to Fable 5.1 across tens of thousands o...

Arena.ai (@arena) — Wed, 09 Sep 2026 20:02:39 GMT

We analyzed how @claudeai's writing has changed from Fable 5 to Fable 5.1 across tens of thousands of high-reasoning Text Arena outputs.

Overall, Fable 5.1 by @AnthropicAI uses fewer agreement openers, fewer em dashes and less wording like “honestly” and “frankly”, while its answers have grown longer.

Compared with Fable 5:
– Agreement openers such as “yes” and “exactly” are 58% less common, appearing in 0.99% of responses versus 2.35%. 
– Honesty wording such as “honestly” and “frankly” falls 45% per 1,000 words. 
– Em dashes fall 32% per 1,000 words.

It has also shifted some patterns in another direction, e.g.: semicolons increase 63% per 1,000 words.

Fable 5.1 still writes long answers, with sentence length and clause frequency close to Opus 5:
– Median response length rises 30% from Fable 5: 319 → 414 words. 
– That remains 21% shorter than Opus 5’s 525 words. 
– Its sentence-length measure is slightly higher than Opus 5’s: 12.54 versus 12.20 words per sentence.

Other familiar patterns also become less common compared with Fable 5:
– Stock phrases such as “load-bearing” fall 20% per 1,000 words. 
– Hedges and caveats such as “perhaps” and “arguably” fall 36% per 1,000 words. 
– Praise and validation appear in 1.98% of responses, down from 3.17%.

Long content words fall from 42.6% to 38.6%, while abstract nouns fall 25%, from 4.39 to 3.28 per 100 words.

Let us know what you think in the comments.

Arena.ai@arena

We analyzed both @claudeai Fable 5 and Fable 5.1 on writing clarity ranked by humans, and are sharing results later today.

Which model do you think writes more clearly?

🔗 View Quoted Tweet

💬30🔄35❤️462👀53591📊72

⚡ Powered by xgo.ing

Game development remains one of the most-requested, and most-challenging, categories on Arena. @iam...

Arena.ai (@arena) — Wed, 09 Sep 2026 17:57:27 GMT

Game development remains one of the most-requested, and most-challenging, categories on Arena.

@iamwaynechi, PhD candidate at Carnegie Mellon University and research intern at Arena, just walked us through GameDevBench: a benchmark built from real tutorials that turns game development into verifiable, deterministic tasks.

How do top frontier models perform on tasks a human beginner could complete in under an hour, and is the biggest bottleneck coding or multimodal understanding?

Check out Wayne’s full talk to hear results and learn more about GameDevBench: youtube.com/watch?v=yoHcvr…

Your browser does not support the video tag.

🔗 View on Twitter

💬10🔄6❤️103👀15924📊22

⚡ Powered by xgo.ing

We analyzed both @claudeai Fable 5 and Fable 5.1 on writing clarity ranked by humans, and are sharin...

Arena.ai (@arena) — Wed, 09 Sep 2026 16:03:04 GMT

We analyzed both @claudeai Fable 5 and Fable 5.1 on writing clarity ranked by humans, and are sharing results later today.

Which model do you think writes more clearly?

💬8🔄3❤️119👀73624📊21

⚡ Powered by xgo.ing

For a limited time, you can test GPT-6 Astra by @OpenAI directly in Arena! Head to Direct Mode, and ...

Arena.ai (@arena) — Wed, 09 Sep 2026 15:17:19 GMT

For a limited time, you can test GPT-6 Astra by @OpenAI directly in Arena! Head to Direct Mode, and select it from the dropdown menu. Find the link below.

GPT-6 Astra (Medium) is available in Direct Mode for the next 24 hours, through Sept. 10 at 8am PT. It will remain available in Battle and Agent Mode after that, as it already is today.

💬39🔄76❤️826👀197860📊176

⚡ Powered by xgo.ing

Use GPT-6 Astra in Direct Mode on Arena now: https://t.co/mErBh5aOmj

Arena.ai (@arena) — Wed, 09 Sep 2026 15:17:19 GMT

Use GPT-6 Astra in Direct Mode on Arena now: bit.ly/3ULCMmj

💬4🔄8❤️82👀13367📊18

⚡ Powered by xgo.ing

See the Code Arena: WebDev Pareto frontier at https://t.co/I584K4VT5z

Arena.ai (@arena) — Tue, 08 Sep 2026 23:16:37 GMT

See the Code Arena: WebDev Pareto frontier at arena.ai/leaderboard/co…

💬0🔄0❤️6👀5633📊1

⚡ Powered by xgo.ing

Muse Spark 1.3 Max by @AIatMeta has reshaped the Pareto frontier for Code Arena: WebDev! Meta's lat...

Arena.ai (@arena) — Tue, 08 Sep 2026 23:16:36 GMT

Muse Spark 1.3 Max by @AIatMeta has reshaped the Pareto frontier for Code Arena: WebDev!

Meta's latest model at Max reasoning is doing something interesting on the Arena Pareto frontier: it's the only model holding down the wide price band between Qwen3.8-max ($5/MToken) and Qwen3.8-Flash-Next ($0.39/MToken). The next closest model to the frontier is Meta's own, Muse Spark 1.3 on xHigh reasoning. 

Muse Spark 1.3 Max performs just 20 pts lower than Qwen3.8 (Max) while costing 30% less, and 24 pts from Kimi K3 (Max) while costing 70% less.
- Muse Spark 1.3 Max: 1650 pts | $3.50/M
- Qwen3.8 (Max): 1670 pts | $5/M
- Kimi K3 (Max): 1674 pts | $12/M

Overall in Code Arena: WebDev, Muse Spark 1.3 Max landed #8, ahead of Claude Fable 5 at #10 (+22 pts), Grok-4.6 (High) at #12 (+26 pts), and GPT-5.6 Sol at #14 (+33 pts).

Congrats to the @AIatMeta team on this release!

Arena.ai@arena

Muse Spark 1.3 (xHigh) just landed @AIatMeta back in the top 10 models for Code Arena: WebDev!

This release is ~#10 in Code Arena: WebDev with 1623 pts (AutoEval). It’s performance is par with Claude Fable 5 at #8 (1628 pts) and GPT-5.6 Sol at #11 (1616 pts).

Muse Spark 1.3 (xHigh) is a huge improvement from previous models Muse Spark 1.1 at #25 (1540 pts), and Muse Spark 1.2 (xHigh) at #30 (1534 pts).

Note: this is an early AutoEval score, in which a Reward Model trained on Arena’s human preference data casts automatic votes in place of live votes. We’ll continue to see how scores converge as more live human votes come in.

Congrats to @alexandr_wang and the @AIatMeta team on this release!

🔗 View Quoted Tweet

💬12🔄35❤️335👀43019📊43

⚡ Powered by xgo.ing

Dive into the scores for the Text-to-Image Arena and more at: https://t.co/G1IeZKsywZ

Arena.ai (@arena) — Tue, 08 Sep 2026 19:03:46 GMT

Dive into the scores for the Text-to-Image Arena and more at: arena.ai/leaderboard/te…

💬0🔄0❤️18👀7499📊2

⚡ Powered by xgo.ing

More exciting news from @OpenAI: GPT-Image-2.5-Sunburst is #1 and Flare is #2 across the Text-to-Ima...

Arena.ai (@arena) — Tue, 08 Sep 2026 19:03:45 GMT

More exciting news from @OpenAI: GPT-Image-2.5-Sunburst is #1 and Flare is #2 across the Text-to-Image Arena, the Image Edit Arena, and the Multi-Image Edit Arena!

Both releases are significant improvements over GPT-Image-2 (medium) in all 3 Arenas. Sunburst delivers the highest quality generation, while Flare still improves upon GPT Image 2 at significantly faster speeds.

The increase by points are (Sunburst | Flare):
- Text-to-Image: +40 | +18 pts
- Single-Image Edit: +59 | +30 pts
- Multi-Image Edit: +81 | +47 pts

Across all categories like 3D Imaging & Modeling, Art and more, Sunburst is #1 and Flare holds #2 in all three Arenas.

Congrats to the team at @OpenAI!

💬33🔄127❤️1412👀207521📊217

⚡ Powered by xgo.ing

Agent Arena measures net improvement relative to the average model across real-world, long-horizon t...

Arena.ai (@arena) — Tue, 08 Sep 2026 17:45:55 GMT

Agent Arena measures net improvement relative to the average model across real-world, long-horizon tasks using web, filesystem, and terminal tools.

See the full leaderboard: arena.ai/leaderboard/ag… and dive into the Agent Arena Pareto frontier: arena.ai/leaderboard/ag…

💬0🔄0❤️8👀4298📊1

⚡ Powered by xgo.ing

GPT-6 Astra (Max) by @OpenAI has reshaped the Pareto frontier for Agent Arena! At $4.01/task for +1...

Arena.ai (@arena) — Tue, 08 Sep 2026 17:45:54 GMT

GPT-6 Astra (Max) by @OpenAI has reshaped the Pareto frontier for Agent Arena!

At $4.01/task for +12.55%, paying about 7% more gets you Claude Fable 5.1 (Max) at $4.30 for a +2pt higher net improvement (+14.5%). Below $4.01, no model on the frontier scores higher than Astra.

💬5🔄7❤️155👀15859📊24

⚡ Powered by xgo.ing

Agent Arena results for GPT-6 Astra (Max) by @OpenAI are in: it debuts at #2 overall with +12.5% net...

Arena.ai (@arena) — Tue, 08 Sep 2026 17:45:53 GMT

Agent Arena results for GPT-6 Astra (Max) by @OpenAI are in: it debuts at #2 overall with +12.5% net improvement across +8.9K real-world agentic sessions!

GPT-6 Astra (Max) scored a massive +40.6% in Praise vs. Complaint (implicit sentiment in user reactions), and +21.9% in Confirmed Success (explicit user feedback that the task worked).

With a $4.01 cost per median task, it also reshapes the Pareto frontier for Agent Arena. More info on Astra’s performance compared to cost below.

Congrats to @OpenAI on this release!

OpenAI@OpenAI

This is GPT-6 Astra.

Anything you can do on a computer, Astra can do for you. Fast.

Your browser does not support the video tag.

🔗 View on Twitter

🔗 View Quoted Tweet

💬33🔄45❤️694👀80552📊89

⚡ Powered by xgo.ing

Check out the Agent Arena Pareto frontier at: https://t.co/Nb8opa5Kkw

Arena.ai (@arena) — Sun, 06 Sep 2026 01:02:47 GMT

Check out the Agent Arena Pareto frontier at: arena.ai/leaderboard/ag…

💬3🔄2❤️15👀12524📊5

⚡ Powered by xgo.ing

Claude Fable 5.1 (Max) by @AnthropicAI has landed in the Agent Arena at #1! At a $4.14 median cost p...

Arena.ai (@arena) — Sun, 06 Sep 2026 01:02:47 GMT

Claude Fable 5.1 (Max) by @AnthropicAI has landed in the Agent Arena at #1! At a $4.14 median cost per task and a +15.8% net improvement, it has reshaped the Pareto frontier! 

Claude Fable 5.1 is both the most performant and costly among all Agent Arena models today.

Based on 6.7k+ real-world agentic sessions, Claude Fable 5.1 stands out in implicit user sentiment with an astonishing (+42.5%) in Praise vs. Complaint. Users are praising it around 2x more often than the next top model. It also sees strong explicit feedback via Confirmed Success (+22.4%), and solid Bash Recovery (+13.1%), with no Tool Hallucinations.

Stay tuned as traces continue to come in for the latest GPT-6 Astra to see how it compares. Use Agent Mode to contribute to the real-world rankings.

💬30🔄31❤️470👀87193📊81

⚡ Powered by xgo.ing

Dig into the details of the Agent Arena leaderboard at: https://t.co/sE9q4FSYAt

Arena.ai (@arena) — Sun, 06 Sep 2026 00:44:04 GMT

Dig into the details of the Agent Arena leaderboard at: arena.ai/leaderboard/ag…

💬0🔄0❤️6👀7148📊1

⚡ Powered by xgo.ing

Claude Fable 5.1 is #1 overall in Agent Arena with +15.8% net improvement, and by signal: - #1 Prais...

Arena.ai (@arena) — Sun, 06 Sep 2026 00:44:03 GMT

Claude Fable 5.1 is #1 overall in Agent Arena with +15.8% net improvement, and by signal:
- #1 Praise vs. Complaint: +42.5%
- #1 Confirmed Success: +22.4%
- #4 Bash Recovery: +13.1%
- #28 Steerability: +0.6%
- No issues with Tool Hallucination

💬4🔄0❤️31👀11762📊7

⚡ Powered by xgo.ing

Claude Fable 5.1 (Max) by @AnthropicAI has landed in the Agent Arena at #1 with +15.8% net improveme...

Arena.ai (@arena) — Sun, 06 Sep 2026 00:44:02 GMT

Claude Fable 5.1 (Max) by @AnthropicAI has landed in the Agent Arena at #1 with +15.8% net improvement across 6.7k+ real-world agentic sessions! It also redraws the price-performance frontier: #1 on the leaderboard at a median cost of $4.14/task.

By signal, Claude Fable 5.1 sees a massive lead in implicit user sentiment with an astonishing (+42.5%) in Praise vs. Complaint. Users are praising it around 2x more often than the next top model. It also sees strong explicit feedback via Confirmed Success (+22.4%), and solid Bash Recovery (+13.1%), with no Tool Hallucinations. More detail on its performance by signal below.

Claude Fable 5.1 (Max) out ranks all past Claude variants and the rest of the pack by a healthy lead.

In Agent Arena, we measure models on millions of long-horizon agentic tasks from a global community of users. Models can access web search, filesystem, and terminal tools to complete complex workflows. The leaderboard measures model performance on outcomes relative to the average model using a causal tracing methodology. 

Stay tuned as traces continue to come in for the latest GPT-6 Astra to see how it compares. Use Agent Mode to contribute to the real-world rankings.

Congrats again to @AnthropicAI for this release.

Claude@claudeai

We’re introducing Claude Fable 5.1 and Claude Mythos 5.1.

They're the world’s most advanced models for coding and knowledge work.

Your browser does not support the video tag.

🔗 View on Twitter

🔗 View Quoted Tweet

💬51🔄34❤️689👀209630📊98

⚡ Powered by xgo.ing

Dive into the Text-to-Video Arena at: https://t.co/IaLvxeRp5d

Arena.ai (@arena) — Sat, 05 Sep 2026 18:23:16 GMT

Dive into the Text-to-Video Arena at: arena.ai/leaderboard/te…

💬1🔄1❤️11👀8646📊3

⚡ Powered by xgo.ing

Grok Imagine Video 1.5 Agent from @SpaceXAI has landed in the Text-to-Video Arena at #5 with 1491 pt...

Arena.ai (@arena) — Sat, 05 Sep 2026 18:23:15 GMT

Grok Imagine Video 1.5 Agent from @SpaceXAI has landed in the Text-to-Video Arena at #5 with 1491 pts!

This release performs on par with Wan-3.0 and FLUX 3 Video (each with 1494 pts), just 3 pts below. Grok Imagine Video 1.5 Agent out performs both Dreamina Seedance-2.5 and 2.0, and MiniMax-H3.

@SpaceXAI is now in the top 5 labs in the Text-to-Video Arena. Congrats to the team!

Grok@grok

Grok Imagine Video 1.5 agent is now available.

Powered by our newest Image 2.0 model, it delivers higher quality, better storytelling from a smarter agent and excels at connecting multiple shots together with greater continuity.

Your browser does not support the video tag.

🔗 View on Twitter

🔗 View Quoted Tweet

💬20🔄22❤️315👀39177📊44

⚡ Powered by xgo.ing

See the Code Arena: WebDev Pareto at https://t.co/a64VZNvZib

Arena.ai (@arena) — Sat, 05 Sep 2026 17:40:42 GMT

See the Code Arena: WebDev Pareto at arena.ai/leaderboard/co…

💬0🔄2❤️19👀10062📊2

⚡ Powered by xgo.ing

GPT-6 Astra (Max) by @OpenAI is #1 on Code Arena: WebDev reshaping the Pareto frontier! It is SOTA ...

Arena.ai (@arena) — Sat, 05 Sep 2026 17:40:41 GMT

GPT-6 Astra (Max) by @OpenAI is #1 on Code Arena: WebDev reshaping the Pareto frontier!

It is SOTA performance at $40/Mtoken, matching the latest Claude model pricing.

In Code Arena, AI models are ranked on web development tasks, including agentic coding workflows that require multi-step reasoning and tool use.

Arena.ai@arena

Real-world results are in. There is a new #1 on Code Arena - GPT-6 Astra (Max)!

It also reshapes the Pareto frontier as the best-performing model at $40/Mtoken, which matches the latest Claude model pricing.

GPT-6 Astra by @OpenAI takes the top spot in Code Arena: WebDev with a score of 1797 pts. This opens up a solid +35pt lead over #2 Claude Fable 5.1 (Max) at 1762 pts and #3 Claude Opus 5 (Max) at 1688 pts.

This is a significant improvement from GPT-5.6 Sol (xHigh) at +180 pts, ranked at #13.

Category level votes still incoming, but already we see it at #1 in: Data & Analytics, Consumer Product, Content Creation Tools and #2 in Gaming and Simulations. 

Stay tuned for other categories like Brand & Marketing, Reference-Based Design and Full Stack rankings.

Congrats to the @OpenAI team on this release!

🔗 View Quoted Tweet

💬40🔄75❤️996👀84546📊122

⚡ Powered by xgo.ing

See the Code Arena: WebDev leaderboard at https://t.co/GFZ3FCC7Cl

Arena.ai (@arena) — Sat, 05 Sep 2026 17:32:42 GMT

See the Code Arena: WebDev leaderboard at arena.ai/leaderboard/co…

💬2🔄1❤️46👀18930📊7

⚡ Powered by xgo.ing

Real-world results are in. There is a new #1 on Code Arena - GPT-6 Astra (Max)! It also reshapes th...

Arena.ai (@arena) — Sat, 05 Sep 2026 17:32:41 GMT

Real-world results are in. There is a new #1 on Code Arena - GPT-6 Astra (Max)!

It also reshapes the Pareto frontier as the best-performing model at $40/Mtoken, which matches the latest Claude model pricing.

GPT-6 Astra by @OpenAI takes the top spot in Code Arena: WebDev with a score of 1797 pts. This opens up a solid +35pt lead over #2 Claude Fable 5.1 (Max) at 1762 pts and #3 Claude Opus 5 (Max) at 1688 pts.

This is a significant improvement from GPT-5.6 Sol (xHigh) at +180 pts, ranked at #13.

Category level votes still incoming, but already we see it at #1 in: Data & Analytics, Consumer Product, Content Creation Tools and #2 in Gaming and Simulations. 

Stay tuned for other categories like Brand & Marketing, Reference-Based Design and Full Stack rankings.

Congrats to the @OpenAI team on this release!

💬177🔄400❤️4196👀796682📊710

⚡ Powered by xgo.ing

Head-to-head: From napkin sketch, to working app using Agent Mode on Arena. Two @arena engineers, G...

Arena.ai (@arena) — Sat, 05 Sep 2026 15:59:05 GMT

Head-to-head: From napkin sketch, to working app using Agent Mode on Arena.

Two @arena engineers, Graham and Logan, both built the same taco truck waitlist app from a single napkin sketch, with a build-vs-buy recommendation on the line.

See who built it better, and try out this prompt yourself with top agents on Agent Arena. 

youtu.be/CVxQS_dFako

Your browser does not support the video tag.

🔗 View on Twitter

💬13🔄4❤️76👀16966📊20

⚡ Powered by xgo.ing

GPT-6 Astra vs. Claude Fable 5.1 vs. Claude Fable 5: Reactions on Cost, Coding & Benchmarks Hea...

Arena.ai (@arena) — Fri, 04 Sep 2026 20:57:58 GMT

GPT-6 Astra vs. Claude Fable 5.1 vs. Claude Fable 5: Reactions on Cost, Coding & Benchmarks

Hear @petergostev’s initial take on GPT-6 Astra by @OpenAI vs Fable 5.1 by @AnthropicAI directly. https://t.co/HU5RPUajc7

💬14🔄7❤️100👀24922📊25

⚡ Powered by xgo.ing

GPT-6 Astra by @OpenAI is now in the Arena! Head to Agent Arena to test it out, and your votes will...

Arena.ai (@arena) — Fri, 04 Sep 2026 20:26:40 GMT

GPT-6 Astra by @OpenAI is now in the Arena!

Head to Agent Arena to test it out, and your votes will shape the evaluation. Scores coming soon.

In Agent Arena, we measure models on millions of real-world, long-horizon agentic tasks. Models can access web search, filesystem, and terminal tools to complete complex workflows. The leaderboard measures model performance on outcomes relative to the average model using a causal tracing methodology.

Also available in Code Arena: WebDev, Text, Vision, Search and Document.

OpenAI@OpenAI

This is GPT-6 Astra.

Anything you can do on a computer, Astra can do for you. Fast.

Your browser does not support the video tag.

🔗 View on Twitter

🔗 View Quoted Tweet

💬19🔄14❤️387👀37898📊45

⚡ Powered by xgo.ing

Test GPT-6 Astra in Battle Mode and Agent Mode at: https://t.co/jFTd7gG8UU

Arena.ai (@arena) — Fri, 04 Sep 2026 20:26:40 GMT

Test GPT-6 Astra in Battle Mode and Agent Mode at: arena.ai

💬2🔄1❤️28👀8680📊5

⚡ Powered by xgo.ing

Test GPT-6 Astra in Battle Mode and Agent Mode at: https://t.co/jFTd7gG8UU

Arena.ai (@arena) — Fri, 04 Sep 2026 20:24:25 GMT

Test GPT-6 Astra in Battle Mode and Agent Mode at: arena.ai

💬0🔄0❤️5👀2892📊1

⚡ Powered by xgo.ing

GPT-6 Astra by @OpenAI is now in the Arena! Head to Agent Arena to test it out, and your votes will...

Arena.ai (@arena) — Fri, 04 Sep 2026 19:21:10 GMT

GPT-6 Astra by @OpenAI is now in the Arena!

Head to Agent Arena to test it out, and your votes will shape the evaluation. Scores coming soon.

In Agent Arena, we measure models on millions of real-world, long-horizon agentic tasks. Models can access web search, filesystem, and terminal tools to complete complex workflows. The leaderboard measures model performance on outcomes relative to the average model using a causal tracing methodology.

Also available in Code Arena: WebDev, Text, Vision, Search and Document.

OpenAI@OpenAI

This is GPT-6 Astra.

Anything you can do on a computer, Astra can do for you. Fast.

Your browser does not support the video tag.

🔗 View on Twitter

🔗 View Quoted Tweet

💬7🔄5❤️118👀8169📊14

⚡ Powered by xgo.ing

Test GPT-6 Astra in Battle Mode and Agent Mode at: https://t.co/jFTd7gG8UU

Arena.ai (@arena) — Fri, 04 Sep 2026 19:21:10 GMT

Test GPT-6 Astra in Battle Mode and Agent Mode at: arena.ai

💬1🔄1❤️11👀2230📊2