StepFun, released May 29, 2026

Step 3.7 Flashprice, context, benchmarks and release details

Provisional: not enough results to rank yet Open weights 13% confidence, 1 of 3 expected sources in

50.3

SI Score

Not ranked yet

Input, per 1M tokens: $0.18
Output, per 1M tokens: $1.11
Context window: 256K
Max output: 256K
Released: May 29, 2026

How the score breaks down

Coding 57.0

Math —

Preference —

Reasoning 47.2

Weights: reasoning 30%, math 15%, coding 40%, preference 15%. Results use fixed 0–100 scales before averaging, and thin evidence is pulled toward 50. Method si-v3-retained-evidence-2, computed Oct 9, 2026, 06:15 UTC.

Around it on the leaderboard

Full leaderboard

Benchmark results

5 benchmarks, 5 results

Each row shows the best published result. Where a model was tested at several settings, such as reasoning effort, open the row to see each one. Hover or tap a value for its source.

Humanity's Last Exam (with tools)reasoning 47.2% 47.2

SWE-bench Procoding 56.3% 56.3

SWE-bench Verifiedcoding 76.5% 76.5

Terminal-Benchcoding 35.6% 35.6

Terminal-Bench 2.1coding 59.6% 59.6

Normalization uses a fixed 0–100 scale for each unit, independent of other models. Compare evaluation conditions before reading a small gap as decisive. “Lab-reported” marks the provider's own published figure.

Details and sources

Open weights: Yes
License: not yet reported
Input modalities: text, image, video
First seen by SuperIndex: Oct 8, 2026
Coverage: 10% of expected source weight

Reported (1)

Official model cards via models.devOct 8, 2026

Awaiting (2)

LiveBench30% of weight
LMArena / Arena60% of weight

Confidence rises as pending sources publish. Some sources never cover some models, so confidence reaches 100% at 80% of expected weight.