Step 5 Preview
StepFun · released
StepFun's new flagship, released as a preview on its apps and API. It scores near the top on long-context reasoning and well on agentic work for its price.
Best for: Low-cost agentic work · Long-context tasks
- Intelligence Index
- 43.7
- #12 of 38
- Price per 1M tokens
- $1 in $2.7 out
- Blended $1.43 at 3:1
- Context window
- 1M
- Up to 64K output
- Output speed
- 89 tokens/s
- $0.72 measured cost per task
Capability profile
One 0–100 score per area, placing this model between the weakest and strongest tracked model across that area's benchmarks.
- Overall64 · #17 of 38
Based on 3 of 3 benchmarks
- Coding59 · #15 of 39
Based on 5 of 5 benchmarks
- Agents & knowledge work59 · #21 of 40
Based on 6 of 6 benchmarks
- Reasoning & science68 · #11 of 39
Based on 2 of 2 benchmarks
- Knowledge & reliability80 · #12 of 39
Based on 2 of 2 benchmarks
- Vision62 · #19 of 31
Based on 2 of 2 benchmarks
Benchmark results
All third-party results on record. Rank is among the tracked models that have been tested on that benchmark.
Overall
#12 of 38· Artificial Analysis
- Vals Index49.4%
#19 of 26· $2.63 per test· Vals AI
- Arena · Text1454 ±11
#20 of 33· 2,701 votes· Arena
Coding
- Terminal-Bench 4.033.3%
#11 of 38· Artificial Analysis
- SciCode58.9%
#8 of 38· Artificial Analysis
- Arena · WebDev1564 ±12
#22 of 34· High effort· 3,138 votes· Arena
- Vibe Code Bench70.4%
#21 of 26· Vals AI
- Code Migration39.5%
#18 of 26· Vals AI
Agents & knowledge work
- GDPval-AA54.2
#15 of 39· Artificial Analysis
- Arena · Agent+0.5 ±1.5
#18 of 29· Arena
- Finance Agent v250.7%
#19 of 26· Vals AI
#19 of 26· Vals AI
- Legal Research Bench40.4%
#16 of 26· Vals AI
- Tax Agent Bench59.5%
#20 of 26· Vals AI
Reasoning & science
- Humanity's Last Exam46.5%
#12 of 39· Artificial Analysis
- CritPt20.9%
#11 of 39· Artificial Analysis
- ARC-AGI-2saturatingNot tested
- ARC-AGI-3Not tested
- GPQA DiamondsaturatingNot tested
Knowledge & reliability
#12 of 39· Artificial Analysis
- AA-LCR88.3%
#2 of 39· Artificial Analysis
Vision
- MMMU-Pro76.4%
#16 of 27· Artificial Analysis
- Arena · Vision1260 ±24
#13 of 19· 604 votes· Arena
Specifications
- Lab
- StepFun
- Released
- Sep 20, 2026
- Availability
- Preview
- Context window
- 1,000,000 tokens
- Max output
- 64,000 tokens
- Accepts
- Text, Images, Video
- Knowledge cutoff
- Not published
- Weights
- Closed
- Parameters
- 600B total · 27B active
- API model ID
step-5-preview
Preview release. StepFun has announced open weights for mid-October; none were published at the time of this snapshot.
Pricing
- Input / 1M tokens
- $1
- Cached input / 1M
- $0.05
- Output / 1M tokens
- $2.7
- Blended (3:1)
- $1.43
- Measured cost / task
- $0.72
Worth knowing
- Specifications come from third-party trackers and the model listings rather than a first-party document.
Official sources
Closest alternatives
The models nearest to this one on overall score.
23 benchmarks tracked. See methodology and sources.