Fireworks(@FireworksAI_HQ)

Partnering up with @arcee_ai to help everyone find a route to own their intelligence is easy when th...

Fireworks (@FireworksAI_HQ) — Thu, 13 Aug 2026 19:18:31 GMT

Partnering up with @arcee_ai to help everyone find a route to own their intelligence is easy when their launch videos serve up good vibes. 

Try our Kimi-K3 on Arcee.

Arcee.ai@arcee_ai

Today we're open-sourcing nac, an agent harness for long-running tasks, and launching the Arcee open models API beta.

Nac is available now under Apache 2.0 on GitHub, built for developers running complex, multi-step engineering workloads.

Your browser does not support the video tag.

💬1🔄1❤️25👀3124📊4

Getting an AI product from prototype to production requires more than choosing a model. This guide...

Fireworks (@FireworksAI_HQ) — Thu, 13 Aug 2026 19:12:09 GMT

Getting an AI product from prototype to production requires more than choosing a model. 

This guide from @Microsoft Teams Calls for Startups breaks down how startups can deploy Fireworks models on Microsoft Foundry and build an architecture designed to evolve as they grow.

Read the full guide: microsoft.com/en-us/startups…

Microsoft for Startups@msft4startups

The model you choose for an #MVP can shape cost, latency, flexibility, and future model choice. 🧩

With @FireworksAI_HQ on#MicrosoftFoundryy, startups can serve open models in Azure and scale toward production with monitoring, caching, governance, and model experimentation.

➡️msft.it/6017aE6Ixz
#MicrosoftForStartupss#GenerativeAII#Developerss

💬0🔄2❤️7👀2089📊2

We tested out Anthropic’s J-Lens probe at two open models: Kimi K3 and Qwen 3.5-9B. By looking at t...

Fireworks (@FireworksAI_HQ) — Thu, 13 Aug 2026 16:56:35 GMT

We tested out Anthropic’s J-Lens probe at two open models: Kimi K3 and Qwen 3.5-9B.

By looking at their internal states, we found that the models are setting up the right vocabulary before a single token hits the page.

Check out the research: fireworks.ai/blog/J-Lens-Ki…

💬1🔄3❤️26👀2380📊8

A lot of attention is on coding agents, but the more interesting work is in the rest of the stack th...

Fireworks (@FireworksAI_HQ) — Wed, 12 Aug 2026 21:32:17 GMT

A lot of attention is on coding agents, but the more interesting work is in the rest of the stack that actually gets them into production and keeps them reliable.

We join @antimetal@braintrust  & @browserbase  for The AI Dev Stack: Beyond Code. 

RSVP: https://t.co/1oBCrdpAd9

💬1🔄2❤️9👀1504📊3

Join Fireworks & @LangChain on August 25th for an exclusive hands-on workshop! Learn to build c...

Fireworks (@FireworksAI_HQ) — Wed, 12 Aug 2026 17:04:25 GMT

Join Fireworks & @LangChain on August 25th for an exclusive hands-on workshop!

Learn to build custom eval models for LangSmith and scale your observability stack to supercharge performance.

After that? Rooftop happy hour.  

Let's go: https://t.co/V86urk4evf

💬0🔄1❤️7👀1564📊2

Qwen3.8-2.4T-A95B is now live on Fireworks with Day-0 support! This 2.4T parameter MoE model is bui...

Fireworks (@FireworksAI_HQ) — Wed, 12 Aug 2026 16:33:10 GMT

Qwen3.8-2.4T-A95B is now live on Fireworks with Day-0 support!

This 2.4T parameter MoE model is built for autonomous agents, heavy coding, large context windows, and is ideal for coding and agentic performance.

Start building today: fireworks.ai/models/firewor…

💬4🔄9❤️120👀29095📊24

Muse Glimmer is live on Fireworks. The new open-weight model from Meta Superintelligence Labs is a ...

Fireworks (@FireworksAI_HQ) — Wed, 12 Aug 2026 01:40:36 GMT

Muse Glimmer is live on Fireworks.

The new open-weight model from Meta Superintelligence Labs is a 30B dense model built for always-on agents that reason across many sequential tool calls and can recover from failures.

128K+ context. Native image and text input.

Start building: fireworks.ai/models/firewor…

💬2🔄2❤️37👀3659📊7

Looking for a faster specialized model for your Agent Work? @nvidia Nemotron 3.5 Lightning (30B MoE...

Fireworks (@FireworksAI_HQ) — Tue, 11 Aug 2026 14:49:20 GMT

Looking for a faster specialized model for your Agent Work?

@nvidia Nemotron 3.5 Lightning (30B MoE, 3B active params) is now live on Fireworks. It’s distilled from NVIDIA Nemotron 3 Ultra to be your high-volume agent engine.

With strong performance on PinchBench and top-tier scores in AA-Omniscience Non-Hallucination, Nemotron 3.5 Lightning is engineered for reliable, high-productivity agentic workflows.

Built for specialization, not generalization. Start building:

app.fireworks.ai/models/firewor…

💬4🔄3❤️47👀3863📊11

Voyage AI models now run natively on Fireworks, the first and only dedicated inference platform @Voy...

Fireworks (@FireworksAI_HQ) — Mon, 10 Aug 2026 20:59:39 GMT

Voyage AI models now run natively on Fireworks, the first and only dedicated inference platform @VoyageAI by @MongoDB has partnered with.

Embed, retrieve, rerank, generate: your full retrieval pipeline on one platform, next to the open models you own.

fireworks.ai/blog/voyage-ai…

💬2🔄1❤️24👀4205📊5

Configure and run Fireworks training jobs from your coding agent. Install the Training Skill in one...

Fireworks (@FireworksAI_HQ) — Mon, 10 Aug 2026 20:07:29 GMT

Configure and run Fireworks training jobs from your coding agent.

Install the Training Skill in one line, describe your goal, and Claude Code, Codex, or Cursor helps choose the method, validate data, estimate cost, launch training, and troubleshoot failures.

fireworksai-docs.mintlify.app/fine-tuning/ag…

💬1🔄1❤️18👀2581📊5

In this episode of @wandb's Gradient Dissent , @l2k and Fireworks CEO @lqiao discuss why she believe...

Fireworks (@FireworksAI_HQ) — Fri, 07 Aug 2026 18:26:20 GMT

In this episode of @wandb's Gradient Dissent , @l2k and Fireworks CEO @lqiao discuss why she believes the industry is at a turning point... 

... one that calls for more open intelligence, not less. 

The full episode is available in the comments.

Weights & Biases@wandb

"We're just at the beginning of this S curve."

says @lqiao, CEO of @FireworksAI_HQ, on this week's Gradient Dissent. 

Fireworks now processes more than 40 trillion tokens a day, more than OpenAI's API and Gemini's, and Lin says usage has expanded fast, from coding last year into legal, finance, recruiting, sales, and a growing wave of co-work tools.

Link to the full episode in the replies.

Your browser does not support the video tag.

💬3🔄1❤️18👀4786📊5

Build in less than a day. Showcase on a global stage. Join us next week @MongoDB's 'The Persistent...

Fireworks (@FireworksAI_HQ) — Thu, 06 Aug 2026 22:18:23 GMT

Build in less than a day.

Showcase on a global stage.

Join us next week @MongoDB's 'The Persistent Context Sprint Hackathon'
- Pier 48, San Francisco, CA 
- Thu, Aug 13 at 12:30 – 8:00 PM

cerebralvalley.ai/e/persistent-c…

Your browser does not support the video tag.

💬4🔄2❤️15👀3209📊7

Build in less than a day. Showcase on a global stage. Join us next week at @MongoDB's 'The Persist...

Fireworks (@FireworksAI_HQ) — Thu, 06 Aug 2026 22:11:02 GMT

Build in less than a day.

Showcase on a global stage.

Join us next week at @MongoDB's 'The Persistent Context Sprint Hackathon'
- Pier 48, San Francisco, CA 
- Thu, Aug 13 at 12:30 – 8:00 PM

cerebralvalley.ai/e/persistent-c…

Your browser does not support the video tag.

💬0🔄0❤️0👀58

This webinar is happening in 30 minutes, and that means there's still time to register! Following t...

Fireworks (@FireworksAI_HQ) — Thu, 06 Aug 2026 17:34:00 GMT

This webinar is happening in 30 minutes, and that means there's still time to register!

Following the discussion will be an open Q&A with our Head of AI Education @Prof_OZ, and the @arizeai team.

See you soon!

Arize AI@arizeai

Next session, August 6 we'll be co leading a discussion with @FireworksAI_HQ:

GPT-5.5, Kimi K3, and 8 More Models: The Real Cost of AI Work

We’ll compare model performance using the metric that matters most: the cost of completing a task successfully.

luma.com/arizeai-RealCo…

💬0🔄1❤️16👀3492📊2

DeepSeek V4 Flash 0731 is ready to fine-tune on Fireworks. Run SFT, DPO, and RL training jobs on th...

Fireworks (@FireworksAI_HQ) — Wed, 05 Aug 2026 20:33:58 GMT

DeepSeek V4 Flash 0731 is ready to fine-tune on Fireworks.

Run SFT, DPO, and RL training jobs on the Dedicated Training API. SFT and DPO from the managed UI. Built for coding agents and high-volume production. 

Strong quality at a fraction of the cost.

docs.fireworks.ai/fine-tuning/mo…

💬14🔄4❤️82👀8935📊29

The future of AI won't be one-size-fits-all. Last week, our President George Hu told @etnshow why w...

Fireworks (@FireworksAI_HQ) — Wed, 05 Aug 2026 00:19:57 GMT

The future of AI won't be one-size-fits-all.

Last week, our President George Hu told @etnshow why we believe specialized intelligence is the next era of enterprise AI.

etn.@etnshow

After signing the open letter in support of open weights Fireworks AI President @GeorgeHuSF joined us to explain why:

"We don't think intelligence should be controlled by just a few frontier labs."

"It should be owned by every company around the world."

"The more open it is, the more access everyone has to intelligence."

"We think the better society and the world is going to be as a whole."

Your browser does not support the video tag.

💬2🔄2❤️23👀3876📊5

If you missed today's K3 on Fireworks webinar, we've got you covered. We'll be partnering with @ari...

Fireworks (@FireworksAI_HQ) — Tue, 04 Aug 2026 22:39:04 GMT

If you missed today's K3 on Fireworks webinar, we've got you covered.

We'll be partnering with @arizeai to chat more models on Thursday, August 6th.

Sign up below!

Arize AI@arizeai

Next session, August 6 we'll be co leading a discussion with @FireworksAI_HQ:

GPT-5.5, Kimi K3, and 8 More Models: The Real Cost of AI Work

We’ll compare model performance using the metric that matters most: the cost of completing a task successfully.

luma.com/arizeai-RealCo…

💬0🔄2❤️17👀4243📊2

ICYMI: This is happening in one hour!

Fireworks (@FireworksAI_HQ) — Tue, 04 Aug 2026 16:00:56 GMT

ICYMI: This is happening in one hour!

Fireworks@FireworksAI_HQ

K3 is here. The world's first open frontier model in the 3-trillion-parameter class. Come see how to run it, train it, and put it to work, on Fireworks. 

Hosted by @jamesr66a  and @Prof_OZ on Tuesday, August 4th, 10am PDT.

Register here: fireworks.ai/event/Webinar-…

💬0🔄0❤️19👀3374📊2

Just over 24 hours left to submit! We've had some cool entries so far. Go go go!

Fireworks (@FireworksAI_HQ) — Mon, 03 Aug 2026 21:23:33 GMT

Just over 24 hours left to submit! 

We've had some cool entries so far.

Go go go!

Fireworks@FireworksAI_HQ

Want to win a Fire Pass?

You have one week to build something that changes your daily life using Kimi K3 on Fireworks.

To enter:
- Tweet a demo video
- Tag us and use #KimiK3onFireworks in your Tweet
- Submit your entry here: forms.gle/yZsJWkecYjPVod…

5 winners get a unique Kimi K3 Fire Pass.

Your browser does not support the video tag.

💬3🔄1❤️30👀5144📊6

https://t.co/VORB78mvnl

Fireworks (@FireworksAI_HQ) — Mon, 03 Aug 2026 20:22:42 GMT

x.com/i/article/2084…

💬7🔄10❤️102👀84774📊41

DeepSeek-V4-Flash-0731 is live on Fireworks, day-zero. DeepSeek reports it beats V4 Pro across all ...

Fireworks (@FireworksAI_HQ) — Sat, 01 Aug 2026 01:51:09 GMT

DeepSeek-V4-Flash-0731 is live on Fireworks, day-zero.

DeepSeek reports it beats V4 Pro across all 9 agentic evals, incl. 82.7% on Terminal Bench.

Better cost-per-task than V4 Pro, at the economical price of $0.28 per 1M with 1M context.

Try it → fireworks.ai/models/deepsee…

💬14🔄23❤️301👀16265📊58

If LoRA is underperforming, don't reach for more expensive full parameter fine-tuning right away. W...

Fireworks (@FireworksAI_HQ) — Fri, 31 Jul 2026 17:12:38 GMT

If LoRA is underperforming, don't reach for more expensive full parameter fine-tuning right away.

We ran three cheap tests (data coverage, optimization, rank) to see if we could close the gap between LoRA and FullFT.

Sometimes we could, sometimes we couldn't.

fireworks.ai/blog/three-tes…

💬1🔄4❤️36👀6445📊10

Fine-tune your own embedding model for the price of a coffee. A great reranker can't surface a doc ...

Fireworks (@FireworksAI_HQ) — Fri, 31 Jul 2026 15:28:29 GMT

Fine-tune your own embedding model for the price of a coffee.

A great reranker can't surface a doc that was never retrieved. A RAG pipeline cannot cite a case it failed to retrieve.

See how contrastive fine-tuning adapts Qwen3-Embedding-8B to specific domains for under $10 on Fireworks. 

Here are 3 straightforward experiments and results, e.g. +36% nDCG.

Read the breakdown → fireworks.ai/blog/fine-tuni…

💬0🔄3❤️27👀3283📊8

Security at the foundation requires openness at the foundation. As open-weight models become critic...

Fireworks (@FireworksAI_HQ) — Fri, 31 Jul 2026 00:33:17 GMT

Security at the foundation requires openness at the foundation.

As open-weight models become critical infrastructure for the next generation of software, the systems around them must be transparent, inspectable, and resilient by design.

That’s why Fireworks is joining @NVIDIA and the Open Secure AI Alliance.

Together, we’re working to ensure the future of AI is not only open, but secure.

NVIDIA@nvidia

Give a big welcome to the newest members of the Open Secure AI Alliance. 💚 
nvda.ws/4pD8Fc53

💬1🔄2❤️26👀3774📊4

Our partners at @depthfirstlabs just released dfs-large1, a specialized model built for finding and ...

Fireworks (@FireworksAI_HQ) — Thu, 30 Jul 2026 18:42:23 GMT

Our partners at @depthfirstlabs just released dfs-large1, a specialized model built for finding and validating real vulnerabilities in large enterprise codebases.

We helped them to scale the training and inference, making this production-capable agent possible. It’s now in preview on the depthfirst platform, and can boast best-in-class performance.

We feel safer already!

Lin Qiao@lqiao

Open weights are a defender's advantage. 

dfs-large1 from @depthfirstlabs matches frontier-model performance on vulnerability discovery and was built on the open GLM-5.2 model, post-trained with RL on @FireworksAI_HQ. 

This is exactly the case for open-source AI policymakers need to notice: American companies can build best-in-class domain models using open weights.

💬3🔄1❤️39👀3483📊7

K3 is here. The world's first open frontier model in the 3-trillion-parameter class. Come see how to...

Fireworks (@FireworksAI_HQ) — Thu, 30 Jul 2026 17:19:04 GMT

K3 is here. The world's first open frontier model in the 3-trillion-parameter class. Come see how to run it, train it, and put it to work, on Fireworks. 

Hosted by @jamesr66a  and @Prof_OZ on Tuesday, August 4th, 10am PDT.

Register here: fireworks.ai/event/Webinar-…

💬2🔄0❤️39👀6540📊8

Weights aren't the only thing that should be open-source. Both kernel repos behind our recent work ...

Fireworks (@FireworksAI_HQ) — Wed, 29 Jul 2026 17:17:55 GMT

Weights aren't the only thing that should be open-source.

Both kernel repos behind our recent work with @MiniMax_AI are now public:
→ Fireworks: github.com/fw-ai/minimax-…
→ Minimax: github.com/MiniMax-AI/MSA…

💬5🔄7❤️86👀41181📊20

DeepSeek V4 Flash isn't just for inference anymore. Fine-tune it on Fireworks with supervised fine-...

Fireworks (@FireworksAI_HQ) — Wed, 29 Jul 2026 02:15:00 GMT

DeepSeek V4 Flash isn't just for inference anymore.

Fine-tune it on Fireworks with supervised fine-tuning, preference tuning, and combined preference optimization from the managed UI. 

Reinforcement learning runs through the Training API. Built for coding agents and high-volume production.

Contact us to get started: fireworks.ai/contact-traini…

💬9🔄4❤️95👀8229📊24

Halfway into 2026, it's clear that the unit economics of AI now matter, especially for security. @...

Fireworks (@FireworksAI_HQ) — Wed, 29 Jul 2026 00:06:52 GMT

Halfway into 2026, it's clear that the unit economics of AI now matter, especially for security. 

@trilogy_ai  built a playbook for cyber defense using Kimi K3 on Fireworks. 

Audit every repo, triage every scan, review every PR.  Get the full details: 

fireworks.ai/blog/trilogy-c…

💬3🔄2❤️14👀2514📊7

You can now fine-tune Kimi K3 on Fireworks. Conduct supervised fine-tuning, preference tuning, and r...

Fireworks (@FireworksAI_HQ) — Tue, 28 Jul 2026 23:22:12 GMT

You can now fine-tune Kimi K3 on Fireworks. Conduct supervised fine-tuning, preference tuning, and reinforcement learning via Training API. 

Run across dedicated, and serverless training. The first open frontier model at 3 trillion parameters, ready for your product.

Contact us to get started: fireworks.ai/contact-traini…

💬2🔄6❤️37👀4255📊8

Want to win a Fire Pass? You have one week to build something that changes your daily life using Ki...

Fireworks (@FireworksAI_HQ) — Tue, 28 Jul 2026 22:39:15 GMT

Want to win a Fire Pass?

You have one week to build something that changes your daily life using Kimi K3 on Fireworks.

To enter:
- Tweet a demo video
- Tag us and use #KimiK3onFireworks in your Tweet
- Submit your entry here: forms.gle/yZsJWkecYjPVod…

5 winners get a unique Kimi K3 Fire Pass.

Your browser does not support the video tag.

💬25🔄14❤️152👀22890📊64

A few ground rules: - Must be built with Kimi K3 on Fireworks, we'll verify before awarding prizes -...

Fireworks (@FireworksAI_HQ) — Tue, 28 Jul 2026 22:39:15 GMT

A few ground rules:
- Must be built with Kimi K3 on Fireworks, we'll verify before awarding prizes
- Substantially AI-generated (not an existing project with AI bolted on)
- One entry per person
- Contest closes Tues Aug 4th at 5pm PT. 

Go!

app.fireworks.ai/models/firewor…

💬4🔄0❤️7👀1634📊5

Many AI tools rent intelligence. Kimi K3 on Fireworks is different. Fine-tune it on your own data,...

Fireworks (@FireworksAI_HQ) — Tue, 28 Jul 2026 18:44:38 GMT

Many AI tools rent intelligence. Kimi K3 on Fireworks is different. 

Fine-tune it on your own data, serve from US-hosted endpoints, and own the weights. Zero data retention. 

ICYMI yesterday, start building today → fireworks.ai/models/firewor…

Your browser does not support the video tag.

💬4🔄4❤️59👀4863📊15

Kimi K3 Day 0 US-hosted Zero Data Retention @cursor_ai 🤝 Fireworks

Fireworks (@FireworksAI_HQ) — Mon, 27 Jul 2026 23:54:43 GMT

Kimi K3 
Day 0
US-hosted
Zero Data Retention

@cursor_ai 🤝 Fireworks

Cursor@cursor_ai

Kimi K3 is now in Cursor! It scores close to the frontier on CursorBench.

It's available on US-based inference thanks to our partners Fireworks, Together, and Baseten. Zero Data Retention is also supported.

💬6🔄18❤️476👀23314📊52

This quote from @brian_armstrong’s viral X post on how @coinbase cut token spend by half while incre...

Fireworks (@FireworksAI_HQ) — Mon, 27 Jul 2026 21:14:32 GMT

This quote from @brian_armstrong’s viral X post on how @coinbase cut token spend by half while increasing token usage clearly stuck with us.

“The goal isn’t to suppress usage. It’s to build the infrastructure that makes exponential growth sustainable.”

That perfectly captures what we’re hearing from hundreds of engineering leaders.

They don’t want to slow AI adoption, they don’t want to tell developers to use AI less. 

They want AI usage to keep growing without AI spend growing at the same rate.

Brian Armstrong@brian_armstrong

How to keep AI spend flat while token usage grows exponentially: Not with friction and spend alerts. With better defaults, routing, and caching.

Better Defaults (not Usage Caps) – Engineers can choose any model they want, but defaults matter. We’re experimenting with defaulting to open weight models like GLM 5.2 and Kimi 2.7 through our LLM gateway, while still encouraging engineers to choose the right model for the task. 91% of our employees were never hitting their usage caps, so instead of lowering caps and driving up alerts, we're moving to cheaper defaults. Note that code reviews use a diversity of models, so they can check each other's work.

Better Routing – In our custom harnesses, we preprocess prompts and route to the best model for the job, considering cache hits and model pricing. For instance, you may want a frontier model for planning, but not for execution where they can be overkill. Ultimately, humans shouldn't be choosing models - AI can automate this task.

Better Caching – Cache misses are the easiest way to drive your cost up. All of our requests are cache aware, so we’re reusing a warm cache wherever possible. For example, our cache hit rate went from 5% → 60% in LibreChat once properly implemented.

Keep Context Lean – Start fresh sessions when switching tasks. Scope file context narrowly. Disconnect unused tools. Don't just compact. The goal isn't fewer tokens used, it's fewer tokens wasted.

Better Visibility – Our engineers can use as many tokens as they want, from whatever model they want, but we’ve made usage visible – and the more you spend on AI, the more impact we expect.

The goal isn't to suppress usage. It's to build the infrastructure that makes exponential growth sustainable.

Putting this into practice has cut our AI spend nearly in half, while our token usage continues to grow.

💬5🔄2❤️27👀8654📊9

The challenge is that building the right infrastructure isn’t trivial. Most teams don’t have the ti...

Fireworks (@FireworksAI_HQ) — Mon, 27 Jul 2026 21:14:32 GMT

The challenge is that building the right infrastructure isn’t trivial.

Most teams don’t have the time to build:
- Intelligent routing that automatically sends routine work to cost-effective open models and reserves frontier models for the hardest reasoning tasks.
- High-performance prompt caching that dramatically reduces redundant token spend.
- Centralized budgets, policies, and visibility across every developer and every AI tool.

That’s why we built Fireworks Nexus.

💬1🔄0❤️1👀1064📊2

Nexus connects to the agentic harnesses your teams already use, whether that’s Claude Code, Codex, O...

Fireworks (@FireworksAI_HQ) — Mon, 27 Jul 2026 21:14:32 GMT

Nexus connects to the agentic harnesses your teams already use, whether that’s Claude Code, Codex, OpenCode, or your own custom tooling.

It gives you:
→ Intelligent routing, automatically matching each task to the right model, from the newly released Kimi K3 and GLM-5.2 to frontier closed models.
→ Enterprise cost controls, with budgets, policies, and usage visibility across your organization.
→ Drop-in compatibility, so your developers don’t have to change the way they work.

The future obviously isn’t fewer tokens. It’s an open intelligence layer that makes exponential token growth economically sustainable.

fireworks.ai/nexus

💬0🔄0❤️2👀955

K3 is here, and Fireworks is first to serve it with zero compromise on quality. Fireworks is the onl...

Fireworks (@FireworksAI_HQ) — Mon, 27 Jul 2026 20:55:39 GMT

K3 is here, and Fireworks is first to serve it with zero compromise on quality. Fireworks is the only provider to match @Kimi_Moonshot quality at launch.

→ Best quality provider on Moonshot's evals
→ 165 tok/s on AA

We're not done optimizing. 

Try it: fireworks.ai/models/firewor…

💬9🔄16❤️220👀17968📊42

https://t.co/9fDMiJMPZt

Fireworks (@FireworksAI_HQ) — Mon, 27 Jul 2026 20:34:38 GMT

x.com/i/article/2081…

💬3🔄3❤️62👀5718📊14

Open models just crossed the intelligence-cost curve. Most of your team's bill is from routine wor...

Fireworks (@FireworksAI_HQ) — Mon, 27 Jul 2026 17:09:07 GMT

Open models just crossed the intelligence-cost curve. 

Most of your team's bill is from routine work at frontier prices. 

We make it easy to own your AI spend, without changing the way your engineers work.

Introducing Fireworks Nexus: fireworks.ai/blog/fireworks…

Your browser does not support the video tag.

💬13🔄11❤️122👀409414📊36

Kimi K3 is now live on Fireworks. Get started → https://t.co/CIP7dJ1OIP

Fireworks (@FireworksAI_HQ) — Mon, 27 Jul 2026 15:38:52 GMT

Kimi K3 is now live on Fireworks.

Get started → fireworks.ai/models/firewor…

Your browser does not support the video tag.

💬10🔄22❤️344👀429035📊64

Start owning your own intelligence today with Kimi K3 on Fireworks. Serve it, fine-tune it, and pu...

Fireworks (@FireworksAI_HQ) — Mon, 27 Jul 2026 15:30:36 GMT

Start owning your own intelligence today with Kimi K3 on Fireworks. 

Serve it, fine-tune it, and put it into production.

What are you waiting for → fireworks.ai/blog/kimik3-on…

💬0🔄1❤️22👀4837📊4

Kimi K3 rivals Anthropic and OpenAI’s top models at a fraction of the cost. $3 / 1M input tokens. $...

Fireworks (@FireworksAI_HQ) — Mon, 27 Jul 2026 15:30:35 GMT

Kimi K3 rivals Anthropic and OpenAI’s top models at a fraction of the cost.

$3 / 1M input tokens.
$15 / 1M output.
$0.30 / 1M cached

You can now route your hardest reasoning to an open model without paying that frontier premium.

💬1🔄1❤️33👀5491📊6

Kimi K3 is live on Fireworks. Day 0, inference and training. US-hosted, and zero data retention. Th...

Fireworks (@FireworksAI_HQ) — Mon, 27 Jul 2026 15:30:34 GMT

Kimi K3 is live on Fireworks. Day 0, inference and training. US-hosted, and zero data retention.

This is the first frontier open model in the 3 trillion parameter class. It sports 1M context, native vision, and reasoning that rivals the top closed models.

Boom.

💬54🔄74❤️1005👀1214292📊220

Kicking off the weekend right at @daytonaio's HackSprint in downtown SF! We've got a table set up w...

Fireworks (@FireworksAI_HQ) — Fri, 24 Jul 2026 18:07:32 GMT

Kicking off the weekend right at @daytonaio's HackSprint in downtown SF!

We've got a table set up with our buddies from @braintrust - stop by and say hello! 

Good luck to all of the hackathon participants, we can't wait to see what you build.

💬1🔄0❤️5👀2148📊2

Many still debate open vs closed, and compare cost per token (accounting metric). Better to shift ...

Fireworks (@FireworksAI_HQ) — Thu, 23 Jul 2026 19:18:33 GMT

Many still debate open vs closed, and compare cost per token (accounting metric). 

Better to shift attention to cost per successful task. @seldo and the team at @arizeai did so across 2,400 runs. 

Conclusion: route by task difficulty, and you win on both cost and coverage.

Arize AI@arizeai

Token price tells you what a model costs to call. It does not tell you what it costs to finish the job.

Arize and @FireworksAI_HQ benchmarked 10 models across 2,400 agent runs, including Kimi K3.

In his tests, Arize's Head of DevRel @seldo found:

• Kimi K3 nearly matched GPT-5.5 overall
• Different models won on different task types
• Routing improved cost and coverage
• Retries and silent failures changed the economics completely

Developers and product managers should optimize for cost per successful task, then use evals and traces to understand where each model belongs.

Get the benchmark data: arize.com/blog/cost-per-…

💬3🔄1❤️31👀5944📊7

@HarryStebbings @lqiao Spotify https://t.co/PeO0QU3NPX Youtube https://t.co/QHDvuh4PbQ Apple Pod...

Fireworks (@FireworksAI_HQ) — Thu, 23 Jul 2026 19:14:59 GMT

@HarryStebbings@lqiao Spotify 
open.spotify.com/episode/14rh37… 
Youtube  
youtu.be/PCAiqKCfRSk?si…
Apple Podcasts 
podcasts.apple.com/us/podcast/20v…

💬0🔄0❤️4👀880📊1

"What does nobody see about the next three years?" - @HarryStebbings "Every single company will ow...

Fireworks (@FireworksAI_HQ) — Thu, 23 Jul 2026 19:14:00 GMT

"What does nobody see about the next three years?" - @HarryStebbings 

"Every single company will own their own intelligence... it's not optional." - @lqiao

Your browser does not support the video tag.

💬3🔄3❤️18👀4273📊7