Arena.ai(@lmarena_ai)

In summary, Jev Router makes smart choices. 4 of its 5 most-used models sit on the Pareto frontier, ...

Arena.ai (@arena) — Wed, 07 Oct 2026 22:29:35 GMT

In summary, Jev Router makes smart choices. 4 of its 5 most-used models sit on the Pareto frontier, and its steerability is notably strong.

But that doesn't yet add up to a better trade-off. Calling DeepSeek V4.1 Flash directly gets similar performance at lower cost and latency.

Our results show why model routing is hard: balancing task success, steerability, cost and latency all at once remains an open challenge.

💬0🔄1❤️7👀3283📊1

We evaluated Jev Router by @typesafeai on Agent Arena. Our tests spanning more than 4,700 real-wor...

Arena.ai (@arena) — Wed, 07 Oct 2026 22:29:33 GMT

We evaluated Jev Router by @typesafeai on Agent Arena. 

Our tests spanning more than 4,700 real-world agentic sessions show:

1. Jev Router does not improve on the current Pareto frontier. For similar performance as DeepSeek V4.1 Flash (Max), the solution costs 38% more and its median model request latency is 1.7x higher.

2. However, it does mostly route to Pareto efficient models. The most LLM it picks is DeepSeek V4.1 Flash, with GPT-6.1 Sol and GPT-6 Luna also being frequent choices.

3. Jev Router’s key strength is steerability. Its score of +10% almost matches Claude Opus 5.5 (High) (+10.48). This highlights the benefit of Jev effectively routing to stronger LLM in response to user feedback.

More insights in the thread below.

💬22🔄10❤️183👀32709📊47

Le Chonk is here, and so are first impressions from @petergostev. Mistral Large 4 by @MistralAI is...

Arena.ai (@arena) — Wed, 07 Oct 2026 19:45:50 GMT

Le Chonk is here, and so are first impressions from @petergostev. 

Mistral Large 4 by @MistralAI is a 1T-parameter model (49B active), with open weights coming at the end of the month.

We put it through a series of real-world tests and compared it with leading proprietary and open models, including GPT-6.1 Sol, Astra, Claude Fable, Claude Opus 5.5, Claude Sonnet 5.5, Kimi K3, GLM 5.3 Flash, DeepSeek Flash and Muse Spark, to help demonstrate where it stands today.

Dive in, and stay tuned for additional scores and updates on Mistral Large 4.

youtube.com/watch?v=YfQ3kE…

💬11🔄4❤️63👀14875📊19

Claude Haiku 5.5 by @AnthropicAI is now in the Arena. Head to Agent Arena to test it out, scores com...

Arena.ai (@arena) — Wed, 07 Oct 2026 18:52:18 GMT

Claude Haiku 5.5 by @AnthropicAI is now in the Arena. Head to Agent Arena to test it out, scores coming soon!

In Agent Arena, we measure models on millions of real-world, 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.

Claude Haiku 5.5 is also available in Code Arena: WebDev, Text, Document and Vision.

Claude@claudeai

Introducing Claude Haiku 5.5: the cheapest, fastest, and most capable small model we’ve ever released.

On average, it costs around 75% less to run than Claude Haiku 4.5.

Your browser does not support the video tag.

💬23🔄9❤️179👀11416📊29

Test Claude Haiku 5.5 in Battle Mode and Agent Mode at: https://t.co/jFTd7gG8UU

Arena.ai (@arena) — Wed, 07 Oct 2026 18:52:18 GMT

Test Claude Haiku 5.5 in Battle Mode and Agent Mode at: arena.ai

💬0🔄0❤️13👀3266📊1

Update: Mistral Large 4’s debut places @MistralAI among the top 15 labs on Code Arena: WebDev! It is...

Arena.ai (@arena) — Wed, 07 Oct 2026 17:05:01 GMT

Update: Mistral Large 4’s debut places @MistralAI among the top 15 labs on Code Arena: WebDev! It is the only European lab to make the list.

Mistral Large 4 is a preview. Its score and ranking may change as @MistralAI continues training the model, and with its open-weights release.

As a preview, this model landed #45 overall with 1534 pts. This is a +304 pt improvement from its previous variant, Mistral Large 3 at #133.

Stay tuned for updated scores, and congrats to @MistralAI on this release!

Arena.ai@arena

Mistral Large 4 by @MistralAI just landed in the Code Arena: WebDev at #45. With 1534 pts, this is a +304 pt improvement from Mistral Large 3 at #130!

This release also marks a 271-point improvement over its next-best-performing variant, Mistral Medium 3.5, at #122.

Mistral Large 4 delivers performance within just 2 points of Claude Opus 4.8 (High) at a nearly 6× lower blended price: $3.47/M vs. $20/M tokens. Nearby but higher-performing models are available for less, keeping it just off the Code Arena: WebDev Pareto frontier.

@MistralAI announced its open weights will be released at the end of October. Stay tuned for its score on Arena within open models, and congrats to the team on this release!

💬17🔄26❤️374👀22364📊46

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

Arena.ai (@arena) — Wed, 07 Oct 2026 06:18:24 GMT

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

💬0🔄1❤️5👀5055📊1

Mistral Large 4 by @MistralAI just landed in the Code Arena: WebDev at #45. With 1534 pts, this is a...

Arena.ai (@arena) — Wed, 07 Oct 2026 06:18:23 GMT

Mistral Large 4 by @MistralAI just landed in the Code Arena: WebDev at #45. With 1534 pts, this is a +304 pt improvement from Mistral Large 3 at #130!

This release also marks a 271-point improvement over its next-best-performing variant, Mistral Medium 3.5, at #122.

Mistral Large 4 delivers performance within just 2 points of Claude Opus 4.8 (High) at a nearly 6× lower blended price: $3.47/M vs. $20/M tokens. Nearby but higher-performing models are available for less, keeping it just off the Code Arena: WebDev Pareto frontier.

@MistralAI announced its open weights will be released at the end of October. Stay tuned for its score on Arena within open models, and congrats to the team on this release!

Mistral AI@MistralAI

Meet Mistral Large 4, aka Le Chonk. 

• 1T parameters, natively multimodal. 49B active.
It is the best open weights model from US or Europe on aggregated benchmarks.
• State-of-the-art on critical workloads, including cyber defense, manufacturing and finance and it surpasses closed frontier models on visual grounding.
• Forged in Europe end-to-end and is deployable from Europe via our own Mistral Cloud infrastructure.
• Available to all via API today. Working with cybersecurity partners privately.

Open weights release end of October.

Your browser does not support the video tag.

💬25🔄22❤️352👀47457📊50

Nano Banana 2.1 by @Google has landed in the Text-to-Image Arena, Image Edit Arena and Multi-Image E...

Arena.ai (@arena) — Tue, 06 Oct 2026 21:01:03 GMT

Nano Banana 2.1 by @Google has landed in the Text-to-Image Arena, Image Edit Arena and Multi-Image Edit Arena! It landed:

- #4 in Multi-Image Edit with 1431 pts
- #5 in Text-to-Image with 1328 pts
- #6 in Image Edit with 1428 pts

This release is a substantial improvement from Nano Banana 2: +80 pts in Text-to-Image, +67 pts in Multi-Image Edit, and +38 pts in Image Edit!

By category, Nano Banana 2.1 stands out in:

- Multi-Image Edit: #3 in Commercial Design and 3D, and #4 across Cartoon/Fantasy, Photorealistic, and Text Rendering
- Text-to-Image: #4 Photorealistic, and #5 in both 3D and Text Rendering
- Image Edit: #4 across Commercial Design, 3D, Cartoon/Fantasy, and Text Rendering

Congrats to the team at @googledeepmind@google on this release!

Google@Google

Meet Nano Banana 2.1, our latest image generation and editing model 🍌

This upgraded version outperforms our previous models across the board, with notable leaps in visual design, mask-based editing, and subject consistency to help you create more natural-looking images.

💬43🔄51❤️653👀64854📊98

Dive into the Multi-Image Edit Arena: https://t.co/j4hWjmnb4z

Arena.ai (@arena) — Tue, 06 Oct 2026 21:01:03 GMT

Dive into the Multi-Image Edit Arena: arena.ai/leaderboard/im…

💬2🔄0❤️8👀5012📊3

Dive into the Multi-Image Edit Arena: https://t.co/j4hWjmnb4z

Arena.ai (@arena) — Tue, 06 Oct 2026 20:24:37 GMT

Dive into the Multi-Image Edit Arena: arena.ai/leaderboard/im…

💬0🔄0❤️2👀941

Nano Banana 2.1 by @Google has landed in the Text-to-Image Arena, Image Edit Arena and Multi-Image E...

Arena.ai (@arena) — Tue, 06 Oct 2026 20:24:36 GMT

Nano Banana 2.1 by @Google has landed in the Text-to-Image Arena, Image Edit Arena and Multi-Image Edit Arena! It landed:

- #4 in Multi-Image Edit with 1431 pts
- #5 in Text-to-Image with 1328 pts
- #6 in Image Edit with 1428 pts

This release is a substantial improvement from Nano Banana 2: +80 pts in Text-to-Image, +67 pts in Multi-Image Edit, and +38 pts in Image Edit!

By category, Nano Banana 2.1 stands out in:

- Multi-Image Edit: #3 in Commercial Design and 3D, and #4 across Cartoon/Fantasy, Photorealistic, and Text Rendering
- Text-to-Image: #4 Photorealistic, and #5 in both 3D and Text Rendering
- Image Edit: #4 across Commercial Design, 3D, Cartoon/Fantasy, and Text Rendering

Congrats to the team at @googledeepmind@google on this release!

Google@Google

Meet Nano Banana 2.1, our latest image generation and editing model 🍌

This upgraded version outperforms our previous models across the board, with notable leaps in visual design, mask-based editing, and subject consistency to help you create more natural-looking images.

💬9🔄6❤️87👀6215📊14

Mistral Large 4 by @MistralAI is now in the Arena! Head to Agent Arena to test it out, and your vot...

Arena.ai (@arena) — Tue, 06 Oct 2026 14:21:52 GMT

Mistral Large 4 by @MistralAI is now in the Arena!

Head to Agent Arena to test it out, and your votes will shape its 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.

Mistral Large 4 is also available in Code Arena: WebDev, Text, and Vision.

Mistral AI@MistralAI

Meet Mistral Large 4, aka Le Chonk. 

• 1T parameters, natively multimodal. 49B active.
It is the best open weights model from US or Europe on aggregated benchmarks.
• State-of-the-art on critical workloads, including cyber defense, manufacturing and finance and it surpasses closed frontier models on visual grounding.
• Forged in Europe end-to-end and is deployable from Europe via our own Mistral Cloud infrastructure.
• Available to all via API today. Working with cybersecurity partners privately.

Open weights release end of October.

Your browser does not support the video tag.

💬11🔄13❤️180👀14113📊23

Test Mistral in Battle Mode and Agent Mode at: https://t.co/jFTd7gG8UU

Arena.ai (@arena) — Tue, 06 Oct 2026 14:21:52 GMT

Test Mistral in Battle Mode and Agent Mode at: arena.ai

💬1🔄0❤️8👀4340📊2

Independent evaluation is essential.

Arena.ai (@arena) — Mon, 05 Oct 2026 20:54:08 GMT

Independent evaluation is essential.

MTS@MTSlive

Arena CEO @ml_angelopoulos argues OpenAI and Anthropic can’t be the only ones deciding whether their own models are safe as agents get powerful enough to break out of sandboxes:

"There's an element of safety that we need to evaluate as well, because given the capabilities of these models to break out of their sandboxes, there needs to be a neutral evaluator for that."

"It's not a matter of if, but when, there's gotta be a neutral evaluation platform, because frankly, the model labs are not incentivized to do it themselves."

"The OpenAIs, the Anthropics of the world, they're passionate about safety, kudos to them. But we need to create a system, and they've called for this as well, that they're not the only ones making the guardrails."
@arena

Your browser does not support the video tag.

💬6🔄1❤️36👀13872📊8

Dive into Agent Arena categories at: https://t.co/sE9q4FSYAt

Arena.ai (@arena) — Mon, 05 Oct 2026 19:02:24 GMT

Dive into Agent Arena categories at: arena.ai/leaderboard/ag…

💬1🔄1❤️8👀4723📊2

The Agent Arena ranks agentic ability to solve problems across domains. U.S. labs lead across ever...

Arena.ai (@arena) — Mon, 05 Oct 2026 19:02:23 GMT

The Agent Arena ranks agentic ability to solve problems across domains. 

U.S. labs lead across every domain:
- @AnthropicAI models hold the #1 position in Code, Work, and Chat
- @OpenAI's GPT 6 Astra (Max) remains in the top four across all three categories: #2 in Code and #4 in both Work and Chat

The leading models stay strong across categories, but their order shifts:
- Claude Fable 5.1 (Max) leads both Code and Work and ranks #2 in Chat
- Claude Sonnet 5.5 (Max) leads Chat and ranks #4 in Code and #5 in Work

Code and Work rankings move together more closely than Chat:
- Gemini 4 Argon (High), for example, ranks #14 in Code and #12 in Work but #5 in Chat, highlighting a distinct strength in conversational agent tasks

Agent Arena ranks models based on overall net improvement against the current frontier, as well as performance across specific categories of agentic tasks. These rankings are grounded in live traces collected from agents attempting to solve real-world tasks submitted by humans around the world.

- Code: writing and debugging code, workflow automation, and data analysis
- Chat: creative writing, learning, personal questions, everyday research, and media generation
- Work: documents, professional research, planning, and professional writing

This visual outlines how selected models compare across Agent Arena categories as of October 5, 2026.

Arena.ai@arena

x.com/i/article/2089…

💬32🔄22❤️389👀37751📊62

See the live results and dive into the Agent Arena Pareto frontier at: https://t.co/5tUiBcBJkO

Arena.ai (@arena) — Fri, 02 Oct 2026 19:48:17 GMT

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

💬0🔄2❤️11👀7377📊2

Exciting news: GPT-6.1 Sol (Max) by @OpenAi just landed in the Agent Arena at #5 (+11.23%) and resha...

Arena.ai (@arena) — Fri, 02 Oct 2026 19:48:16 GMT

Exciting news: GPT-6.1 Sol (Max) by @OpenAi just landed in the Agent Arena at #5 (+11.23%) and reshaped the Pareto frontier!

At a $0.56 median cost per task, it delivers performance within 2 percentage points of GPT-6 Sol and GPT-6 Astra for substantially less cost:
- 39% lower cost than GPT-6 Sol, while scoring +1.52 pts higher
- 81% lower cost than GPT-6 Astra, while landing within 1.04 pts

GPT-6.1 Sol also delivers top-five performance at substantially lower cost compared to:
- 88% lower cost than Claude Fable 5.1 (Max), while landing within 3.08 pts (ranked #1)
- 65% lower cost than Claude Opus 5.5 (High), while landing within 2.59 pts (ranked #2)
- 80% lower cost than Claude Sonnet 5.5 (Max), while landing within 1.29 pts (ranked #3)

Congrats to the team @OpenAI on this release!

💬39🔄35❤️428👀58615📊85

See the live results and dive into the Agent Arena Pareto frontier at: https://t.co/5tUiBcBJkO

Arena.ai (@arena) — Fri, 02 Oct 2026 19:33:52 GMT

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

💬0🔄0❤️4👀5892📊1

Claude Sonnet 5.5 by @AnthropicAI just landed at #3 in the Agent Arena. This model has a median cost...

Arena.ai (@arena) — Fri, 02 Oct 2026 19:33:51 GMT

Claude Sonnet 5.5 by @AnthropicAI just landed at #3 in the Agent Arena. This model has a median cost per task of $2.74, and a +12.5% net improvement score.

Claude Sonnet 5.5 delivers top-tier performance, but at a cost premium: #2 Claude Opus 5.5 costs $1.58 per task while achieving a higher score. That tradeoff keeps Sonnet 5.5 just off the Agent Arena Pareto frontier.

💬23🔄17❤️331👀40166📊57

Dive into the Code Arena: WebDev leaderboard at https://t.co/CY0WQfSRRg

Arena.ai (@arena) — Fri, 02 Oct 2026 19:31:31 GMT

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

💬1🔄1❤️6👀5194📊2

Exciting news: Claude Sonnet 5.5 (Max) by @AnthropicAI has debuted at #3 in the Agent Arena with +12...

Arena.ai (@arena) — Fri, 02 Oct 2026 19:31:30 GMT

Exciting news: Claude Sonnet 5.5 (Max) by @AnthropicAI has debuted at #3 in the Agent Arena with +12.5% net improvement!

This release is a 8.1 percentage-point increase over Claude Sonnet 5 (High), which ranks #13 with +4.4% net improvement. By category, Claude Sonnet 5.5 secured the #1 spot in Chat (+15.6%) above both Fable 5.1 (+11.49%) and Opus 5.5 (+10.29%).

This performance comes with a higher cost: Claude Sonnet 5.5 (Max) has a median cost of $2.74 per task, about 73% higher than #2 Claude Opus 5.5 (High) at $1.58.

@AnthropicAI models now hold all three top positions in Agent Arena. Congrats to the team!

💬26🔄26❤️450👀60153📊73

This Week in the Arena: Four releases reshaped the Pareto frontier across Text, Code, and Agent Aren...

Arena.ai (@arena) — Fri, 02 Oct 2026 17:24:13 GMT

This Week in the Arena: Four releases reshaped the Pareto frontier across Text, Code, and Agent Arena.

- After OpenAI’s DevDay, GPT-6.1 Sol (Max) entered Code Arena: WebDev at #3 and joined the Pareto frontier at $8/MToken.

- However, once Sonnet 5.5 entered the leaderboard later that day, Sol moved to #4 and fell off the Pareto frontier. Today, Sonnet is 2 points behind #2 GPT-6 Astra (Max) at 80% lower cost.

- Excitingly, Gemini 4 Argon (High) landed at #1 in Text Arena, #8 in Code Arena: WebDev (1679 pts), and #8 in Agent Arena (+7.92% net improvement).

- Finally, MiMo-V2.6-Pro and Flash entered Agent Arena at #5 and #9 among open-source models, respectively.

Agent scores for GPT-6.1 Sol and Sonnet 5.5 are coming soon. Stay tuned! 
 
See leaderboards, other updates, and more in less than 90 seconds in the video. 

youtube.com/watch?v=Mh0buA…

💬36🔄4❤️182👀29446📊50

How to design rewards for post-training frontier image models? Our research suggests human prefere...

Arena.ai (@arena) — Fri, 02 Oct 2026 15:05:12 GMT

How to design rewards for post-training frontier image models? 

Our research suggests human preference reward is necessary, but insufficient: A preference model may still reward outputs that look appealing but miss details, introduce unrequested content, or exhibit other forms of reward-hacking.

We therefore optimize towards a composite reward:
- Bradley-Terry reward model trained on ~5.6M pairwise human votes
- Faithfulness reward from auto-generated prompt checklists evaluated by a vision-language model
- Constraint reward covering explicit and implicit user intent
- Anti-reward-hacking rubric rewards targeting failures such as garbled text and photorealism drift
This post-training recipe improves two already-strong open image models:
- Post-trained FLUX.2-dev gains 69 Elo points on our live T2I leaderboard, scoring 1202
- Post-trained Ideogram 4 gains 20 Elo points reaching a score of 1224 and surpassing all publicly listed open models (as of Sep 04, 2026).

Offline ablations with Gemini 3.5 Flash as the judge, show that these reward components are complementary: win rate against the base model increases as we add faithfulness and then constraint rewards on top of preference-only training, reaching 64.2%.

Finally, we ensemble policies trained with and without the anti-reward-hacking objective directly in weight space, further increasing win rate to 66.0%.

💬17🔄13❤️158👀20748📊37

MiMo-V2.6-Flash has landed on the Agent Arena Pareto frontier with -0.57% net improvement at a $0.04...

Arena.ai (@arena) — Thu, 01 Oct 2026 18:57:59 GMT

MiMo-V2.6-Flash has landed on the Agent Arena Pareto frontier with -0.57% net improvement at a $0.04 median cost per task!

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

💬4🔄4❤️83👀15755📊13

MiMo-V2.6-Pro and MiMo-V2.6-Flash by @XiaomiMiMo have landed in Agent Arena! MiMo-V2.6-Pro ranks #5...

Arena.ai (@arena) — Thu, 01 Oct 2026 18:57:58 GMT

MiMo-V2.6-Pro and MiMo-V2.6-Flash by @XiaomiMiMo have landed in Agent Arena!

MiMo-V2.6-Pro ranks #5 among open-source models with +3.17% net improvement across 8.1K+ real-world agentic sessions. This release is a +10.4 percentage-point lift and a 9-spot improvement from MiMo-V2.5-Pro at #13 with -7.23%!

Its strongest signal is Confirmed Success (explicit user feedback that the task worked) where it scored +7.35% and ranks #2 among open-source models!

MiMo-V2.6-Flash landed #9 among open models with -0.57% net improvement across 13K+ real-world agentic sessions. 

At a $0.04 median cost per task (56% lower than MiMo-V2.6-Pro), it also landed on the Agent Arena Pareto frontier! See its placement below.

Congrats to the @XiaomiMiMo team on the release!

Xiaomi MiMo@XiaomiMiMo

Introducing Xiaomi MiMo-V2.6 — Pro & Flash.

Frontier intelligence, all the modalities, built in public.

🔹 Two omnimodal models, advancing through scaled reinforcement learning
🔹 Pro performs on par with Claude Opus 5 and GPT-5.6 Sol across most agent benchmarks
🔹 Pro scores 46 on the Artificial Analysis Intelligence Index — the highest among open-source models
🔹 Stronger coding, computer use, 3D reasoning and creative capabilities
🔹 Open model weights, technical report, RL environments and training code

mimo.xiaomi.com/mimo-v2-69Ymvj

💬29🔄18❤️368👀41420📊52

Dive into the Code Arena: WebDev leaderboard at https://t.co/CY0WQfSRRg

Arena.ai (@arena) — Thu, 01 Oct 2026 16:51:03 GMT

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

💬1🔄1❤️8👀5060📊2

Exciting update: Claude Sonnet 5.5 with xHigh reasoning has landed in the Code Arena: WebDev. With 1...

Arena.ai (@arena) — Thu, 01 Oct 2026 16:51:02 GMT

Exciting update: Claude Sonnet 5.5 with xHigh reasoning has landed in the Code Arena: WebDev. With 1786 pts, its ranked #3!

At a blended $8/M tokens, Claude Sonnet 5.5 remains on the Pareto frontier with xHigh reasoning. This release is just 2 pts from GPT-6 Astra in the #2 spot with 1788 pts, for 80% of the price.

By domain, Claude Sonnet 5.5 (xHigh) landed:
- #2 in Gaming, Reference-Based Design, and Brand & Marketing
- #3 in Simulations
- #4 in Content Creation Tools and Consumer Product
- #6 in Data & Analytics

Congrats again to @AnthropicAI on this release!

💬44🔄23❤️506👀87923📊93

Arena Open House: Rooftop Happy Hour 10/7. We're throwing open the doors to our new SF HQ office an...

Arena.ai (@arena) — Thu, 01 Oct 2026 14:42:18 GMT

Arena Open House: Rooftop Happy Hour 10/7.

We're throwing open the doors to our new SF HQ office and want to invite researchers, developers, and builders pushing on hard AI problems up to the roof! Come hang out, meet the team, and check out the space - Bay Area views included.

Boba bar, light bites, and refreshments provided.
Space is limited. Register to request a spot on the list!
luma.com/8jmuq1f2

💬3🔄2❤️34👀10827📊7

Gemini 4 Argon by @GoogleDeepMind just went head-to-head with top frontier models across a set of ge...

Arena.ai (@arena) — Thu, 01 Oct 2026 00:09:01 GMT

Gemini 4 Argon by @GoogleDeepMind just went head-to-head with top frontier models across a set of generations.

At 33% lower cost per task than GPT-6.1 Sol, and 70% lower than Claude Opus 5.5, see first impressions from @petergostev on how Gemini 4 Argon compares.

youtube.com/watch?v=h5EL5z…

💬26🔄15❤️309👀32439📊58

Hidream-O1-Video-1.0 by @HiDream_AI just landed in the Image-to-Video Arena at #7 with 1456 pts! Th...

Arena.ai (@arena) — Wed, 30 Sep 2026 22:32:34 GMT

Hidream-O1-Video-1.0 by @HiDream_AI just landed in the Image-to-Video Arena at #7 with 1456 pts!

This model is available on @vivago_ai, is 8 pts away from gemini-omni-flash and within 20 pts of both dreamina-seedance-2.0 and 2.5.

Congrats to @HiDream_AI on this release!

vivago.ai (HiDream)@vivago_ai

New HiDream models just landed in vivago R1 Studio 🚀

Introducing:
• HiDream-O1 Image 2.0
• HiDream-O1 Editing 1.5
• HiDream-O1 Video

All three models are now available in vivago R1 Studio - bringing the latest HiDream image generation, editing, and video capabilities directly into your creative workflow.

New models. New possibilities. Go make something the internet can’t ignore. 🔥

💬8🔄5❤️107👀21919📊17

Dive into the Image-to-Video leaderboard at: https://t.co/dAaKcypuH6

Arena.ai (@arena) — Wed, 30 Sep 2026 22:32:34 GMT

Dive into the Image-to-Video leaderboard at: arena.ai/leaderboard/im…

💬0🔄0❤️8👀5802📊1

Gemini 4 Argon (High) has reshaped the Pareto frontier with a $0.62 cost per task! Dive into the Ag...

Arena.ai (@arena) — Wed, 30 Sep 2026 21:35:39 GMT

Gemini 4 Argon (High) has reshaped the Pareto frontier with a $0.62 cost per task!

Dive into the Agent Arena details at arena.ai/leaderboard/ag…

💬1🔄2❤️40👀13979📊7

More big news from @GoogleDeepMind: Gemini 4 Argon (High) is #8 in Agent Arena with a +7.92% net imp...

Arena.ai (@arena) — Wed, 30 Sep 2026 21:35:38 GMT

More big news from @GoogleDeepMind: Gemini 4 Argon (High) is #8 in Agent Arena with a +7.92% net improvement score, and has reshaped the Pareto frontier with a $0.62 cost per task! See its placement below.

Gemini 4 Argon (High) is a 4.96 percentage point improvement over Gemini 3.8 Flash (High), at #19 with +2.96% net improvement.

By key signals, Gemini 4 Argon (High) stands out in:
- #1 in Steerability with +15.88% (the model’s ability to course-correct when you push back)
- #2 in Confirmed Success with +14.15% (explicit user feedback that the task worked)
- #4 in Praise vs Complaint with +27.74% (implicit sentiment in user reactions)

By category Gemini 4 Argon (High) is especially strong in Chat, landing at #3 with +11.58% net improvement.

With 3k real-world agentic sessions so far, this score is preliminary. Stay tuned as more traces come in from our global community of users.

Congrats to the @GoogleDeepMind team on this release!

Arena.ai@arena

Big news: Gemini 4 Argon (High) by @GoogleDeepMind just landed #1 in Text Arena with 1525 pts, and #8 in Code Arena: WebDev with 1679 pts!

This release has reshaped the Text Arena Pareto frontier with a blended $8/MToken! Gemini 4 Argon (High) is now the most cost efficient model, see its placement on Pareto frontier below.

In the Text Arena, Gemini 4 Argon (High) ranks #1 in Coding, Hard Prompts, Instruction Following, Longer Query, and Creative Writing. It also leads every occupational domain evaluated, with additional #1 spots in English, Non-English, Chinese, and Russian.

This model is +20 points above the #2 ranked Claude Opus 4.6 (High), and a huge leap from Google’s previous release, Gemini 3.8 Flash (High) at #11!

In Code Arena: WebDev, Gemini 4 Argon (High) gained +96 points from Gemini 3.8 Flash (High), and went from #29 to #8.

Congrats to the @GoogleDeepMind team on this impressive frontier release!

💬18🔄31❤️583👀62459📊66

Dive into the Text Arena leaderboard at https://t.co/q0J9NAytLa and the Code Arena: WebDev leaderboa...

Arena.ai (@arena) — Wed, 30 Sep 2026 20:30:25 GMT

Dive into the Text Arena leaderboard at arena.ai/leaderboard/te… and the Code Arena: WebDev leaderboard at arena.ai/leaderboard/co…

💬1🔄0❤️25👀16848📊4

With a blended $8/MToken and 1525 pt score, Gemini 4 Argon (High) reshaped the Text Arena Pareto fro...

Arena.ai (@arena) — Wed, 30 Sep 2026 20:30:25 GMT

With a blended $8/MToken and 1525 pt score, Gemini 4 Argon (High) reshaped the Text Arena Pareto frontier!

💬1🔄12❤️171👀29028📊22

Big news: Gemini 4 Argon (High) by @GoogleDeepMind just landed #1 in Text Arena with 1525 pts, and #...

Arena.ai (@arena) — Wed, 30 Sep 2026 20:30:24 GMT

Big news: Gemini 4 Argon (High) by @GoogleDeepMind just landed #1 in Text Arena with 1525 pts, and #8 in Code Arena: WebDev with 1679 pts!

This release has reshaped the Text Arena Pareto frontier with a blended $8/MToken! Gemini 4 Argon (High) is now the most cost efficient model, see its placement on Pareto frontier below.

In the Text Arena, Gemini 4 Argon (High) ranks #1 in Coding, Hard Prompts, Instruction Following, Longer Query, and Creative Writing. It also leads every occupational domain evaluated, with additional #1 spots in English, Non-English, Chinese, and Russian.

This model is +20 points above the #2 ranked Claude Opus 4.6 (High), and a huge leap from Google’s previous release, Gemini 3.8 Flash (High) at #11!

In Code Arena: WebDev, Gemini 4 Argon (High) gained +96 points from Gemini 3.8 Flash (High), and went from #29 to #8.

Congrats to the @GoogleDeepMind team on this impressive frontier release!

Google DeepMind@GoogleDeepMind

Introducing Gemini 4 Argon – our new frontier model.

It’s built for complex workflows across coding, enterprise knowledge work, and cybersecurity defense – rolling out today to a set of trusted testers through our Fairwind Program.

💬139🔄323❤️3695👀879431📊504

Dive into the Text Arena leaderboard at https://t.co/q0J9NAytLa and the Code Arena: WebDev leaderboa...

Arena.ai (@arena) — Wed, 30 Sep 2026 20:28:13 GMT

Dive into the Text Arena leaderboard at arena.ai/leaderboard/te… and the Code Arena: WebDev leaderboard at arena.ai/leaderboard/co…

💬0🔄0❤️2👀2681

With a blended $8/MToken and 1525 pt score, Gemini 4 Argon (High) reshaped the Text Arena Pareto fro...

Arena.ai (@arena) — Wed, 30 Sep 2026 20:28:13 GMT

With a blended $8/MToken and 1525 pt score, Gemini 4 Argon (High) reshaped the Text Arena Pareto frontier!

💬2🔄3❤️27👀4608📊6

Can Claude Sonnet 5.5 and GPT-6.1 Sol deliver strong results at a fraction of what Fable, Opus or As...

Arena.ai (@arena) — Wed, 30 Sep 2026 19:40:57 GMT

Can Claude Sonnet 5.5 and GPT-6.1 Sol deliver strong results at a fraction of what Fable, Opus or Astra would cost?

We just tested them side-by-side on multiple prompts, tracking what each generation actually costs, and comparing results to larger models from their labs.

See the results from @petergostev on our YouTube.

youtu.be/r0ymhRtcTeI?si…

💬17🔄2❤️144👀20644📊30

At a blended $8/MToken ($2 input / $10 output per 1M tokens), GPT-6.1 Sol reshaped the Pareto fronti...

Arena.ai (@arena) — Wed, 30 Sep 2026 18:42:05 GMT

At a blended $8/MToken ($2 input / $10 output per 1M tokens), GPT-6.1 Sol reshaped the Pareto frontier!

Dive into the Code Arena: WebDev leaderboard and Pareto frontier at arena.ai/leaderboard/co…

💬0🔄1❤️20👀4682📊3

Exciting news: GPT-6.1 Sol (Max) by @OpenAI just landed the Code Arena: WebDev at #3 with 1759 pts, ...

Arena.ai (@arena) — Wed, 30 Sep 2026 18:42:04 GMT

Exciting news: GPT-6.1 Sol (Max) by @OpenAI just landed the Code Arena: WebDev at #3 with 1759 pts, and at a blended $8/MToken it reshapes the Pareto frontier!

GPT-6.1 Sol (Max) marks a clear improvement in cost efficiency: it gained 70 points over GPT-6 Sol (Max) for the same price.

It landed within 30 points of GPT-6 Astra (Max) at 80% lower blended token cost, and 59 points from Claude Opus 5.5 (Max) at 50% of the price. See position on the Pareto frontier for the Code Arena: WebDev in the post below.

Overall, GPT-6.1 Sol improved from GPT-6 Sol by 4 rankings! It also improved in every category:
- Consumer Product: #5 → #1
- Simulations: #6 → #3
- Data & Analytics: #4 → #3
- Content Creation Tools: #4 → #3
- Gaming: #6 → #4
- Reference-Based Design: #6 → #4
- Brand & Marketing: #10 → #6

Congrats to the @OpenAI team on the release!

💬23🔄38❤️624👀115338📊93

Dive into the Image Edit leaderboard at: https://t.co/ubjCDMJOqh

Arena.ai (@arena) — Wed, 30 Sep 2026 16:38:04 GMT

Dive into the Image Edit leaderboard at: arena.ai/leaderboard/im…

💬1🔄0❤️7👀3512📊2

Exciting news: Ideogram 4.5 by @ideogram_ai has entered the Image Edit Arena’s top 20! This release...

Arena.ai (@arena) — Wed, 30 Sep 2026 16:38:03 GMT

Exciting news: Ideogram 4.5 by @ideogram_ai has entered the Image Edit Arena’s top 20!

This release scored 1351 pts, and ranking #18 overall. Ideogram 4.5 landed within the top 20 for the following Image Edit categories:
- #14 in 3D Imaging & Modeling 
- #15 in Product & Commercial Design
- #17 across Cartoon, Anime & Fantasy; Photorealistic & Cinematic; and Text Rendering
- #18 in Portraits
Ranking for Art coming soon.

Ideogram's release notes highlight that this launch focuses on a unique capability for multi-turn edits existing pixels are persevered. Check out more details in their post.

Congrats to the @ideogram_ai team on this release!

Ideogram@ideogram_ai

Introducing Ideogram 4.5, the most precise edit model.

With each edit, leading models add artifacts, pixel shifts, and color changes. Ideogram 4.5 eliminates artifact buildup, making multi-turn editing possible.

Live in Ideogram, the API, and launch partners. Open weights soon.

Your browser does not support the video tag.

💬10🔄15❤️169👀20625📊25

Use Claude Sonnet 5.5 in Direct Mode on Arena now: https://t.co/vS2BYZNmxK

Arena.ai (@arena) — Wed, 30 Sep 2026 14:58:16 GMT

Use Claude Sonnet 5.5 in Direct Mode on Arena now: bit.ly/4hjJNUn

💬0🔄1❤️12👀3916📊2

Exciting update: you can now test Claude Sonnet 5.5 by @AnthropicAI directly in Arena for a limited ...

Arena.ai (@arena) — Wed, 30 Sep 2026 14:58:15 GMT

Exciting update: you can now test Claude Sonnet 5.5 by @AnthropicAI directly in Arena for a limited time! Head to Direct Mode, and select it from the dropdown menu. Find the link below.

Claude Sonnet 5.5 (High) is available in Direct Mode for the next 48 hours, through Oct. 2 at 8am PT. It will remain available in Battle and Agent Mode after that, as it already is today.

Claude@claudeai

Introducing Claude Sonnet 5.5, the second model in the Claude 5.5 family.

It’s a clear upgrade over Sonnet 5, runs more than 30% faster, and costs up to 30% less for most work.

Your browser does not support the video tag.

💬12🔄10❤️207👀27897📊29