Step 5 Preview benchmarks
Step 5 Preview has 6 published scores in this dataset, from a StepFun release dated 2026-09-20. Its composite index is 78.7, which places it 17th of 22 ranked models and 12th among 15 open-weight models. Its strongest area is agentic (index 96.3) and its weakest is coding (index 47.2). Its best placing is on GPQA Diamond, where its 93.5 ranks 7th of 19 models with a published score.
Category profile
Index 0–100: each score relative to the best published score on that test.
Model facts
- Lab
- StepFun
- Type
- Open weights
- Released
- 2026-09-20
- Parameters
- 600B total, 27B active (MoE)
- Context
- 1M
- License
- StepFun Model License 1.0 (Apache-2.0-style with extra conditions)
Step 5 Preview hardware requirements and cost
Even at 4-bit it needs about 414 GB (4× H200 (564 GB)), well beyond a home workstation, so most people run it on rented multi-GPU servers. Full BF16 precision needs about 1.4 TB (Multiple nodes (over 1.1 TB)).
| Precision | Weights | With headroom | Smallest setup |
|---|---|---|---|
| BF16 (full precision) | 1.2 TB | 1.4 TB | Multiple nodes (over 1.1 TB) |
| FP8 / 8-bit | 600 GB | 690 GB | 8× H200 (1,128 GB) |
| 4-bit (GGUF Q4_K_M) | 360 GB | 414 GB | 4× H200 (564 GB) |
Estimate: parameters × bytes per parameter, plus 15% for the KV cache and runtime at modest context lengths. Long contexts and large batches need more.
Lighter or stronger open models
Open models ranked just above and below on the composite index, with their 4-bit memory needs.
- GLM 5.382.7520 GB
- DeepSeek V4.1 Flash79.8381 GB
- Gemma 4 31B79.221 GB
- Nemotron 3 Ultra74.4379 GB
- Gemma 4 26B A4B72.917 GB
- gpt-oss-120b57.381 GB
Step 5 Preview speed and pricing
Measured figures link to their source; estimates are labelled.
- API output speed
- 146 tok/sArtificial Analysis median · Artificial Analysis
- Time to first token
- 2.97 sOpenRouter model page provider stats
- API price per 1M tokens
- $1.00 in · $2.70 outlab's own endpoint on OpenRouter · OpenRouter endpoints API
Score by test
Rank is among every model with a published score on that test.
Nearby in the composite ranking
Models with the closest composite index.
Head-to-head comparisons
Side-by-side pages built from the tests both models report.
No head-to-head page for Step 5 Preview yet. Compare it on the rankings dashboard.
Sources
| Model | Test | Score | Setting | Source |
|---|---|---|---|---|
| Step 5 Preview | GPQA Diamond | 93.5 | High effort (vendor-reported) | cellcog.ai (relaying StepFun's release table) |
| Step 5 Preview | Humanity's Last Exam | 46.5 | no tools, High effort (vendor-reported) | cellcog.ai (relaying StepFun's release table) |
| Step 5 Preview | Terminal-Bench 4.0 | 33.3 | Terminal-Bench v4, High effort (vendor-reported) | cellcog.ai (relaying StepFun's release table) |
| Step 5 Preview | BrowseComp | 88.7 | High effort (vendor-reported) | cellcog.ai (relaying StepFun's release table) |
| Step 5 Preview | MMMU-Pro | 76.0 | MMMU-Pro (not original MMMU), High effort (vendor-reported) | cellcog.ai (relaying StepFun's release table) |
| Step 5 Preview | LMArena Text Elo | 1448 | Step 5 Preview, ±6, board dated 2026-10-27 | LMArena (arena.ai) text leaderboard |
Scores are published results from model cards, lab reports, and public leaderboards, each linked to its source. Settings such as thinking mode or test version differ between sources and are shown next to each score. OpenLLM Buddy catalog models are uncensored fine-tunes; their scores are those of the official base model. Data collected 2026-10-30. How the index works