Scale-up possibilities for everyone. API Platform: https://xitter.twitterapi.xyz/t.co/XlaiCy4DGm HuggingFace: https://xitter.twitterapi.xyz/t.co/isMgbv7XSm
Joined February 2025
- Tweets695
- Following181
- Followers16.6K
- Likes1.4K
Pinned Tweet
Introducing Step 5 Preview: Advancing the Pareto Frontier.
Step 5 Preview is our new flagship model for agentic work, delivering frontier-level performance across software engineering and professional knowledge work, with particular strength in finance.
- 600B total / 27B active MoE, with 1M context + Vision
- Substantially lower task cost at comparable intelligence
- Broad software engineering capabilities with sustained execution over long horizons
Try Step 5 Preview: platform.stepfun.ai
Model page: stepfun.com/step-5-preview
Open weights on Oct 15.
How do we build AI agents that run longer, perform reliably, and deliver better results—all while keeping costs under control?
Join @StepFun_ai, @Tiny_Fish, @FeatherlessAI, @MonidHQ , and @MatrAIx2026 at Agent Night Out to explore the challenges of building reliable, long-running agents in production.
Nov 17 | 6–9 PM | SF
luma.com/stepfun-624j?tk=FYT…
Sign up and meet us: luma.com/stepfun-624j?tk=FYT…
Great to be at AI Engineer Shanghai 2026!
Thanks to everyone who stopped by our booth! It was great connecting with developers, sharing live demos, and exchanging ideas on long-horizon agents and the future of agentic AI.
Our very own Ailing Teng @EileenTal also took the stage to share how Step 5 Preview and our open-source Agent Harness, Step Code, work together to help agents stay on track and reliably execute long-horizon tasks.
We really appreciated all the thoughtful conversations throughout the event.
Stay tuned for what we're building next!
Rain couldn't stop the builders.
Great conversations at our Step 5 Preview Launch Night with @togethercompute!
From long-horizon agents and tool-calling reliability to inference costs and what AI breakthroughs mean for the next wave of startups.
Huge thanks to our panelist Dan Fu (@realDanFu · @togethercompute), Mar Hershenson (@MarHershenson · @pearvc), Saoud Rizwan (@sdrzn · @cline), and Joey Wang (@harvey), and everyone who joined us!
More to come.
StepFun retweeted
sf, join us this thursday for step 5 preview launch night with stepfun
we’ll dig into the model, building + serving agent workloads, and what’s worth building as models keep improving, with dan fu, hongyu zhou, mar hershenson, saoud rizwan, joey wang
rsvp: luma.com/12t5e4lc
What a weekend at LA Hacks 2026!
Beyond connecting with the community at our booth, we hosted a hands-on workshop to help participants get started with Step 5 Preview and explore what they could build with it.
We also launched the StepFun Challenge and saw teams made especially thoughtful and effective use of our models.
Huge thanks to everyone who stopped by, joined our workshop, or took on the challenge. Congrats to our three winning teams!
We loved being part of LA Hacks and can't wait to see what you build next.
Step 5 Preview weights are out. Now let’s meet in SF.🍷
Join @StepFun_ai and @togethercompute for Step 5 Preview Launch Night & AI Founders Meetup on Thu, Oct 22.
On stage: Hongyu Zhou (@StepFun_ai), Dan Fu (@realDanFu · @togethercompute), Mar Hershenson (@MarHershenson · @pearvc), Saoud Rizwan (@sdrzn · @cline), and Joey Wang (@harvey).
A first look at Step 5 Preview, live Q&A, and a panel on what's still worth building as AI models get better.
Food, drinks & great conversations.
A first look at Step 5 Preview, live Q&A, and a panel on what's still worth building as AI models get better. Food, drinks & great conversations.
Want to join us? Request your spot here: luma.com/12t5e4lchttps://lum…
StepFun retweeted
congrats to the StepFun team on releasing Step 5 Preview open weights today
we’re excited to be a launch partner and to support another StepFun release as they bring Step 5 Preview to developers
Step 5 Preview is now open weights.
BF16 and FP8 weights are live on Hugging Face and ModelScope, together with our technical blog.
• Sparse MoE: 600B total, 27B active per token
• 1M-token context window
• Native text and vision input
• Built for long-horizon agentic work: coding, software engineering, professional knowledge work and finance
Frontier intelligence, open weights. Build what's next.
Hugging Face: huggingface.co/stepfun-ai/St…
ModelScope: modelscope.ai/models/stepfun…
GitHub: github.com/stepfun-ai/Step-5…
A 1M-token context window, endless possibilities.
Excited to see Step 5 Preview live on @DeepInfra! Thanks to the team for making it easier for developers to build with Step 5.
What will you build first?
Step 5 Preview from @StepFun_ai is live on DeepInfra, day zero.
600B open-weight MoE, 27B active. 1M context, text + vision. FP8.
$1.00 in / $2.70 out / $0.05 cached per 1M tokens.
Built for long-horizon agents: coding, frontend, finance.
What would you load into 1M tokens first?
Thanks to @radixark for helping power Step 5 Preview in production!
Great to see SGLang enabling efficient, scalable inference for real-world workloads. Excited to keep pushing the boundaries together.
Congratulations to @StepFun_ai on Step 5 Preview going open weights! RadixArk is proud to help power StepFun's first-party production API with SGLang at scale.
Production inference means great token economics under strict latency requirements. SGLang and its cache-aware routing stack now bring that to Step 5 Preview in proven production traffic.
Building with Step 5 Preview? Deploy with SGLang today!
StepFun retweeted
Congratulations to @StepFun_ai on Step 5 Preview going open weights! RadixArk is proud to help power StepFun's first-party production API with SGLang at scale.
Production inference means great token economics under strict latency requirements. SGLang and its cache-aware routing stack now bring that to Step 5 Preview in proven production traffic.
Building with Step 5 Preview? Deploy with SGLang today!
Step 5 Preview is now open weights.
BF16 and FP8 weights are live on Hugging Face and ModelScope, together with our technical blog.
• Sparse MoE: 600B total, 27B active per token
• 1M-token context window
• Native text and vision input
• Built for long-horizon agentic work: coding, software engineering, professional knowledge work and finance
Frontier intelligence, open weights. Build what's next.
Hugging Face: huggingface.co/stepfun-ai/St…
ModelScope: modelscope.ai/models/stepfun…
GitHub: github.com/stepfun-ai/Step-5…
StepFun retweeted
Step 5 Preview: The new pareto frontier of agentic finance is here
We ran an experiment: which open model gives the best M&A memo per dollar?
We had 6 open models write the same memo for 80 companies, and two other AI judges picked the best
xitter.twitterapi.xyz/StepFun_ai/status/2110…
Step 5 Preview is now open weights.
BF16 and FP8 weights are live on Hugging Face and ModelScope, together with our technical blog.
• Sparse MoE: 600B total, 27B active per token
• 1M-token context window
• Native text and vision input
• Built for long-horizon agentic work: coding, software engineering, professional knowledge work and finance
Frontier intelligence, open weights. Build what's next.
Hugging Face: huggingface.co/stepfun-ai/St…
ModelScope: modelscope.ai/models/stepfun…
GitHub: github.com/stepfun-ai/Step-5…
Great to see Step 5 Preview live on @FeatherlessAI!
Frontier-level intelligence, competitive cost efficiency, and easier access for developers.
Step 5 Preview is here.
@StepFun_ai's new open-weights model sets a new Pareto frontier, matching Kimi K3 on the @ArtificialAnlys Intelligence Index!
It’s built for long-horizon agent work, with frontier-level results on software engineering and professional knowledge work, especially finance.
- 600B sparse MoE, 27B active per token
- ~1/6 the cost per task of Claude Opus 5.5, per Artificial Analysis
Now on Featherless, served at FP8.
featherless.ai/model-pages/s…
Great to see Step 5 Preview delivering strong inference performance on NVIDIA GB300.
Looking forward to seeing what developers build with it.
Congrats to StepFun on Step 5 Preview 🙌 Their largest open weights release.
The 600B MoE model is built to handle agent tasks lasting up to 24 hours.
The model runs at nearly 8K tokens/sec/GPU on NVIDIA GB300 using NVFP4.
NVP4 checkpoint: huggingface.co/stepfun-ai/St…
Deploy with @SGL_project - docs.sglang.io/cookbook/auto… or @vLLM_project - recipes.vllm.ai/stepfun-ai/S…
Post-train with NeMo Automodel: github.com/NVIDIA-NeMo/Autom…
Run it with NVIDIA NIM using a single command: catalog.ngc.nvidia.com/orgs/…
Step 5 Preview is now open weights.
BF16 and FP8 weights are live on Hugging Face and ModelScope, together with our technical blog.
• Sparse MoE: 600B total, 27B active per token
• 1M-token context window
• Native text and vision input
• Built for long-horizon agentic work: coding, software engineering, professional knowledge work and finance
Frontier intelligence, open weights. Build what's next.
Hugging Face: huggingface.co/stepfun-ai/St…
ModelScope: modelscope.ai/models/stepfun…
GitHub: github.com/stepfun-ai/Step-5…
Our technical blog, "Inside Step 5 Preview: Architecture and Training for Agentic Work," covers how architecture and training were co-designed for longer, more complex agent tasks:
• 92-layer narrow-and-deep architecture for implicit multi-hop reasoning
• Sparse GQA to cut million-token context cost, with 8-token block indexing reducing indexing and top-k overhead to 1/8
• FP32 precision on key states plus fine-grained gradient clipping for stable training
• Million-scale verifiable environments
• RL for long-horizon tasks with context compression built into policy training
platform.stepfun.ai/blog/ins…
We've also updated Step Code to v0.1.4, our open-source agent harness built around Step 5 Preview.
It helps agents keep track of the goal on long, complex tasks, keep making progress, and check against verifiable results whether the job is actually done.
Four releases since open-sourcing, about one every 3 working days, mostly driven by community feedback. Thank you to everyone who tried it and shared feedback.
github.com/stepfun-ai/Step-C…
