Step 3.7 FlashvsStep 3.5 Flash
Across 6 shared benchmarks, Step 3.7 Flash leads overall: Step 3.7 Flash wins 4, Step 3.5 Flash wins 2, with 0 ties and an average score difference of +2.05.
Step 3.7 Flash
StepFunAI · 2026-05-29 · Reasoning model
Step 3.5 Flash
StepFunAI · 2026-02-02 · Chat model
Step 3.7 Flash4 wins(67%)(33%)2 winsStep 3.5 Flash
Benchmark scores
Grouped by capability, sorted by largest gap within each. 6 shared benchmarks.
Agent Level Benchmark
Step 3.7 Flash 2/2| Benchmark | Step 3.7 Flash | Step 3.5 Flash | Diff |
|---|---|---|---|
| τ²-Bench - Telecom | 98.504 / 264Thinking (With Tools) | 94.4035 / 264Thinking (With Tools) | +4.10 |
| Terminal Bench Hard | 35.6070 / 244Thinking (With Tools) | 32.6093 / 244Thinking (With Tools) | +3 |
AI Agent - Information Search
Step 3.7 Flash 1/1| Benchmark | Step 3.7 Flash | Step 3.5 Flash | Diff |
|---|---|---|---|
| BrowseComp | 75.8227 / 57Thinking (With Tools) | 6931 / 57Thinking (With Tools) | +6.82 |
General Evaluation
Step 3.5 Flash 1/1| Benchmark | Step 3.7 Flash | Step 3.5 Flash | Diff |
|---|---|---|---|
| GPQA Diamond | 80.90225 / 462Thinking (No Tools) | 83.10191 / 462Thinking (No Tools) | -2.20 |
General Knowledge
Step 3.5 Flash 1/1| Benchmark | Step 3.7 Flash | Step 3.5 Flash | Diff |
|---|---|---|---|
| CritPt | 2.30126 / 200Thinking (No Tools) | 2.50125 / 200Thinking (No Tools) | -0.20 |
Instruction Following
Step 3.7 Flash 1/1| Benchmark | Step 3.7 Flash | Step 3.5 Flash | Diff |
|---|---|---|---|
| IF Bench | 67.3090 / 282Thinking (No Tools) | 66.5093 / 282Thinking (No Tools) | +0.80 |
Specs
| Field | Step 3.7 Flash | Step 3.5 Flash |
|---|---|---|
| Publisher | StepFunAI | StepFunAI |
| Release date | 2026-05-29 | 2026-02-02 |
| Model type | Reasoning model | Chat model |
| Architecture | MoE | MoE |
| Parameters | 198B | 196B |
| Context length | 256K | 256K |
| Max output | Not available | 16K |
API pricing
Prices use DataLearner records when available; missing fields are not inferred.
| Item | Step 3.7 Flash | Step 3.5 Flash |
|---|---|---|
| Text input | ¥1.35 / 1M tokens | Not public |
| Text output | ¥8.1 / 1M tokens | Not public |
| Cache read | ¥0.27 / 1M tokens | Not public |
One or both models have incomplete public pricing.
Summary
- Step 3.7 Flashleads in:Agent Level Benchmark (2/2), AI Agent - Information Search (1/1), Instruction Following (1/1)
- Step 3.5 Flashleads in:General Evaluation (1/1), General Knowledge (1/1)
On average across the 6 shared benchmarks, Step 3.7 Flash scores 2.05 higher.
Largest single-benchmark gap: BrowseComp — Step 3.7 Flash 75.82 vs Step 3.5 Flash 69 (+6.82).
Page generated from structured model, pricing and benchmark records. No real-time LLM is used to write the prose.