StepFun's Step 5 Preview Matches Top Rivals at One-Third the Cost
StepFun's new 600B MoE flagship matches Kimi K3 on the Intelligence Index at roughly one third the cost, with open weights due in October.
- StepFun released Step 5 Preview, a 600B total, 27B active MoE reasoning model with a 1M context window.
- Scores 44 on the Artificial Analysis Intelligence Index, matching Kimi K3 (max) and one point below GLM-5.3 and Qwen3.8 Max.
- Priced at $1 input and $2.70 output per 1M tokens, about 2.8x cheaper per task than same-tier rivals.
- Frontier reasoning is the strength: 46% on Humanity's Last Exam and 21% on CritPt physics.
- Weakness is agentic work: trails peers on GDPval-AA, Terminal-Bench 4.0 and AA-Briefcase, plus a 43% hallucination rate.
- Available on StepFun's API now, with open weights release scheduled for October 15th.
Step 5 Preview reaches a 44 benchmark score at lower API prices
Chinese AI lab StepFun has released Step 5 Preview through its first-party API. The flagship reasoning model succeeds Step 3.7 Flash, and StepFun says open weights will follow on October 15.
Artificial Analysis gives Step 5 Preview a score of 44 on its composite Intelligence Index, compared with a median of 24 among comparable models. That ties Kimi K3 (max) and falls one point behind GLM-5.3 (max) and Qwen3.8 Max. Its listed API prices are substantially lower than those peers.
600B parameters, 27B active
| Specification | Step 5 Preview |
|---|---|
| Architecture | 600B total parameters; 27B active mixture-of-experts |
| Context window | 1 million tokens |
| Input | Text and images; StepFun also lists video support |
| Output | Text |
| Input price | $1 per 1 million tokens |
| Output price | $2.70 per 1 million tokens |
| Cache discount | 95%, according to StepFun |
| Generation speed | 92.8 tokens per second on StepFun’s API |
| Availability | API available; open weights scheduled for October 15 |
A mixture-of-experts model routes each token through part of the network rather than activating every parameter. The 27B active figure approximates per-token computation. The full 600B count remains relevant to storage, memory, and distributed serving requirements.
A 44 score at roughly one-third the task cost
Artificial Analysis reports spending $924.68 to run Step 5 Preview across its Intelligence Index, equivalent to about $0.72 per task. Kimi K3 (max) posts the same score at approximately $2 per task, and GLM-5.3 (max) costs about $2.01 per task for a score of 45.
| Model | Intelligence Index | Cost per task |
|---|---|---|
| Step 5 Preview | 44 | $0.72 |
| Kimi K3 (max) | 44 | $2.00 |
| GLM-5.3 (max) | 45 | $2.01 |
Step 5 generated roughly 164 million output tokens during the evaluation, close to Kimi K3’s 170 million. Per-token rates account for most of the cost gap: Kimi charges $3 per million input tokens and $15 per million output tokens, compared with StepFun’s $1 and $2.70. MiMo-V2.6-Pro is the cited exception, combining a higher index score with a lower cost of $0.13 per task.
Production costs will vary with prompt length, output length, cache reuse, concurrency, and provider discounts. The benchmark figures offer a controlled comparison rather than a forecast for every application.
Reasoning drives the score
Step 5 Preview’s largest gains over Step 3.7 Flash appear on demanding reasoning tests. It scores 46% on Humanity’s Last Exam, an expert-level multidisciplinary benchmark, and 21% on CritPt, which tests physics reasoning. Those results improve on its predecessor by 25 and 19 percentage points, respectively, and exceed the reported scores for GLM-5.3 (max) and Qwen3.8 Max on both evaluations.
Broad factual knowledge also contributes to the composite score. Step 5 reaches 42% accuracy on AA-Omniscience, ahead of the 753B-parameter GLM-5.3 (max) at 34% and behind the 2.8T-parameter Kimi K3 (max) at 48%. Parameter comparisons require caution because training data, architecture, and active computation differ across models.
Agents expose the gap
Evaluations involving tools, terminal work, office-style tasks, and multi-step automation produce weaker relative results. Step 5 trails Qwen3.8 Max and GLM-5.3 (max) across all four agent-oriented benchmarks cited by Artificial Analysis.
| Benchmark | Step 5 Preview | Qwen3.8 Max | GLM-5.3 (max) |
|---|---|---|---|
| GDPval-AA | 1,566 Elo | 1,668 Elo | 1,646 Elo |
| Terminal-Bench 4.0 | 33% | 39% | 42% |
| AA-Briefcase | 1,432 Elo | 1,640 Elo | 1,525 Elo |
| AutomationBench-AA | 51% | 56% | 62% |
Higher scores are better across these tests. Percentage results represent benchmark success rates, and Elo scores are relative ratings that should be compared only within the same evaluation.
Frequent answers bring factual risk
Artificial Analysis reports that Step 5 attempts 68% of AA-Omniscience questions and hallucinates on 43% of those attempts. GLM-5.3 answers 55% of questions and records a 30% hallucination rate among its attempts. Step 5 consequently reaches 16 on the combined AA-Omniscience Index, only two points above GLM-5.3’s 14 despite its higher raw accuracy.
Applications that publish factual answers without review should evaluate the model on domain-specific data and add retrieval, citations, confidence thresholds, or human approval. The reported tendency to answer rather than abstain makes raw accuracy an incomplete reliability measure.
Reasoning jobs fit; agents lag
| Workload | Practical assessment |
|---|---|
| Cost-sensitive reasoning and research | Strong candidate based on its reasoning scores and listed token prices. |
| Scientific problem solving | Promising CritPt and Humanity’s Last Exam results support further testing. |
| Long-document analysis | The 1 million-token limit provides capacity, but teams should measure retrieval and reasoning quality at their target lengths. |
| Tool-using and terminal agents | Qwen3.8 Max and GLM-5.3 (max) lead on the cited agent benchmarks. |
| Unsupervised factual publishing | The hallucination rate calls for stronger validation and abstention controls. |
| Image and video workflows | StepFun lists support, but the cited results do not establish multimodal quality. |
The first-party API produced 92.8 output tokens per second in Artificial Analysis testing. End-to-end latency will also depend on time to first token, prompt size, concurrency, network conditions, and provider load.
Open weights bring deployment questions
Until the scheduled weights arrive, Step 5 Preview remains primarily an API product. A self-hosted deployment will depend on the release license, checkpoint formats, quantization support, routing implementation, and available inference tooling.
The model’s 27B active parameters reduce computation per token, but serving infrastructure still needs to store or distribute the 600B-parameter network. Self-hosted economics will therefore depend on memory capacity, interconnects, batching, and utilization rather than StepFun’s API prices alone.
Comments
