@FireworksAI_HQ

The frontier platform for training and inference on open-weights models at scale.

fireworks.ai
Joined September 2022
  • Tweets1.7K
  • Following307
  • Followers34.5K
529Media
Today we're launching the Specialized Intelligence Index (SII): one destination for real-work benchmarks across industries, built by the teams that use them every day. Hear from Fireworks co-founder @the_bunny_chen on the importance of specialized benchmarks:
19
35
23
175
57,583
Think a smart chatbot can help your family find a new home? Think again. @JoshPix, the SVP of AI at @Zillow, is coming to Fireworks Forge to share how Zillow trains its own models for one of life's big decisions. Join us Nov 3 in SF: bit.ly/forgex
1
9
2,004
Fireworks retweeted
Fireworks
@FireworksAI_HQ16h
DeepSeek V4.1 Flash 1017 runs at 1,521+ tok/s on Fireworks Ultrafast. Fast enough to ship Tetris in 11 seconds, 7x faster than Claude Opus 5.5. Join the Ultrafast waitlist: bit.ly/3UA0wKq
80
163
89
2,215
265,311
We trained @Meta’s Muse Glimmer on Kalshi outcomes. One pass put it 11.2 percentage points ahead of Jev in prediction accuracy. In a backtest, it returned +7.1% to Jev’s −8.1%. Good thing the money was hypothetical. Train it for free with SFT through Nov 30. Try our recipe: fireworks.ai/blog/train-muse…
3
7
7
70
4,958
DeepSeek V4.1 Flash 1017 runs at 1,521+ tok/s on Fireworks Ultrafast. Fast enough to ship Tetris in 11 seconds, 7x faster than Claude Opus 5.5. Join the Ultrafast waitlist: bit.ly/3UA0wKq
80
163
89
2,215
265,311
Guten Morgen, Berlin! Our team is at GTC Berlin ready to answer your questions about how open models can perform at frontier levels. Today, join Fireworks, @Microsoft & @NVIDIA for a fireside chat on when prompting or routing is enough, and the path from open model to production. Save your spot: luma.com/fireworks-tony
Image of a promotional graphic for NVIDIA GTC. Features the Brandenburg Gate at sunset with a geometric overlay. Text reads: "Build AI. Open Possibility. It all starts here. October 20–22 | Berlin." NVIDIA GTC logo is at the bottom left.

ALT Image of a promotional graphic for NVIDIA GTC. Features the Brandenburg Gate at sunset with a geometric overlay. Text reads: "Build AI. Open Possibility. It all starts here. October 20–22 | Berlin." NVIDIA GTC logo is at the bottom left.

4
2
10
1,933
Muse Spark Open is available Day-0 on Fireworks, for inference and training. One of the most capable US open weights models to date: built in the US by @AIatMeta, served on US-only infrastructure with zero data retention. Try it now: fireworks.ai/models/firework…
Bar chart titled "Muse Spark Open on Rogo’s Big Finance Bench" comparing AI performance. Three categories: Easy, Medium, Hard. In "Easy," Muse Spark Open leads with 88.1%, followed by GPT-4 at 88.7%, and Claude Opus 5.5 at 84.5%. In "Medium," Muse Spark Open scores 44.0%, GPT-4 at 41.3%, and Claude Opus 5.5 at 30.7%. In "Hard," Muse Spark Open at 17.3%, GPT-4 at 16.5%, and Claude Opus 5.5 at 6.6%. Muse Spark Open generally has higher scores. Data sourced from Fireworks' Specialized Intelligence Index.

ALT Bar chart titled "Muse Spark Open on Rogo’s Big Finance Bench" comparing AI performance. Three categories: Easy, Medium, Hard. In "Easy," Muse Spark Open leads with 88.1%, followed by GPT-4 at 88.7%, and Claude Opus 5.5 at 84.5%. In "Medium," Muse Spark Open scores 44.0%, GPT-4 at 41.3%, and Claude Opus 5.5 at 30.7%. In "Hard," Muse Spark Open at 17.3%, GPT-4 at 16.5%, and Claude Opus 5.5 at 6.6%. Muse Spark Open generally has higher scores. Data sourced from Fireworks' Specialized Intelligence Index.

2
4
25
2,908
On Nov 2, join us w/ @Microsoft x @nvidia for an SF buildathon that pits frontier models vs specialized open models. To win: Pick a task, build an eval & beat a frontier model on quality, latency or cost. Teams of 2-5 compete for $10K. Apply early to save your spot: luma.com/fireworks-8kmc
1
1
4
2,481
We’ve identified a security incident involving unauthorized access to certain Fireworks platform functionality, affecting a limited subset of customers. Please see our security bulletin
Fireworks
@FireworksAI_HQOct 10

Fireworks October 2026 security incident

Article · @FireworksAI_HQ
4
1
4
39
12,985

Fireworks October 2026 security incident

Article · @FireworksAI_HQ
3
1
8
49
22,441
Ever given an agent permissions in production, then wondered what it would do next? @YonatanOren_, Founding Researcher at @ArmadinSecurity, shares how his team post-trains open models to monitor security agents at Fireworks Forge. Join us Nov 3 in SF: bit.ly/forgex
1
1
2
5
2,858
Our VP of AI spoke recently at @AIconference on how to harness frontier models. Check out his keynote below:
Rob Ferguson
@RobFergusonIIIOct 9
In 2024 I made 5-year AI predictions on stage at the AI Conference. This year I went back and graded myself.My TAIC 2026 keynote, “Your Own Frontier”: youtube.com/watch?v=LKfD599x… @AIconference @FireworksAI_HQ
2
2
1
18
3,242
What if agents remembered things in their weights as well as their context? @realJessyLin, Co-founder of @EngramLab, speaks at Fireworks Forge on building agents with native memory. Join us Nov 3 in SF: bit.ly/forgex
2
1
8
2,243
What are you spinning first at the analog party?
Tech Week
@Techweek_Oct 9
Vinyl records, tacos, fresh coconuts and craft cocktails at Harlan Records. All analog, all night. @fin_ai, @TrustVanta, @FireworksAI_HQ, @linear, @DustHQ and District came together to put a different kind of spin on #SFTechWeek 🎶
4
1,612
We have been having such a fun time at #SFTechWeek! Thanks to the @Techweek_ team for being such great partners!
Tech Week
@Techweek_Oct 9
the people building some of SF's most online products, spending a night analog. Thanks @fin and @FireworksAI_HQ for all the incredible events you hosted this week! #SFTechWeek @fin_ai @TrustVanta @linear @FireworksAI_HQ @DustHQ @alanmcglinchey
5
1,836
Ever wonder why so many teams are building on open models now? @natolambert has been writing about it for years. The co-founder of @trillium_labs and author of Interconnects is speaking at Forge on how open models win. Join us Nov 3 in SF: bit.ly/forgex
1
1
11
3,917
Tokenomics, from the chip all the way up. @dylan522p, Founder & CEO of @SemiAnalysis_, joins the Fireworks Forge speaker list on optimizing every layer of AI infrastructure for inference at scale. Join us Nov 3 in SF: bit.ly/forgex
1
1
9
2,326
Just another example of how companies are bringing specialized intelligence to their customers with the help of Fireworks. Congrats on the launch of Athena Inference Model (AIM) @ZetaGlobal !
David A Steinberg
@dsteinberg10000Oct 8
We just made some of the biggest announcements in @ZetaGlobal's history. Introducing AthenaOS, our enterprise intelligence operating system that starts with what you want to achieve, and Athena MCP, bringing Zeta intelligence into the tools businesses already use. Day one integrations with ChatGPT, Claude, Palantir and Gemini. We also unveiled our new Athena Inference Model (AIM) being developed with @FireworksAI_HQ using @nvidia Nemotron, and announced early work with partners on a specialized chip through a capital-light approach.e What a massive kickoff to #ZetaLive26! $zeta LG 🚀🚀🚀 Read more here: businesswire.com/news/home/2…
2
1
15
2,115
Fireworks retweeted
Sophia Yang
@sophiamyangOct 8
Lights. Cameras. Evals. 🎬 @swyx takes the Forge stage to break down domain-specific everything: data, models, evals, chips. Join us!
🤖 Made with AI
Fireworks
@FireworksAI_HQOct 8
Every serious AI team eventually builds their own data, evals, models, and maybe even chips. @swyx of @latentspacepod and @aidotengineer on why the future is domain-specific everything. Join us Nov 3 in SF: bit.ly/3TxgUuB
2
3
27
6,758
Load more