SI Index Data as of Oct 9, 2026

Step 3.7 Flash: Price, Context, Benchmarks, and Release Details

stepfun provisional listing open weights · first seen Oct 8, 2026 · score computed Oct 9, 2026, 01:50 UTC
50.3
SI Score (method si-v3-retained-evidence-2)
13% confidence 13 percent, Low confidence — 1 of 3 expected sources in
Coverage 10% of expected source weight · 100% confidence at 80% coverage

Pillar breakdown

Coding (weight 40 percent) 57.0
Math (weight 15 percent) —
Preference (weight 15 percent) —
Reasoning (weight 30 percent) 47.2

Pillar weights: reasoning 30% · math 15% · coding 40% · preference 15%. Benchmark results use fixed 0–100 scales before averaging; incomplete evidence is shrunk toward 50.

Facts

Input price / 1M
$0.18models.devFirst-party hosted API; MIT models.dev transcription. Provider documentation: https://platform.stepfun.com/docs/zh/overview/concept. Exact canonical endpoint; lowest short-context Standard USD token tier; cache/batch discounts excluded. Deprecated endpoints excluded; stepfun/step-3.7-flash Retrieved Oct 9, 2026 · MIT
Open source ↗
Output price / 1M
$1.11models.devFirst-party hosted API; MIT models.dev transcription. Provider documentation: https://platform.stepfun.com/docs/zh/overview/concept. Exact canonical endpoint; lowest short-context Standard USD token tier; cache/batch discounts excluded. Deprecated endpoints excluded; stepfun/step-3.7-flash Retrieved Oct 9, 2026 · MIT
Open source ↗
Context window
256Kmodels.devPublished source fact Retrieved Oct 9, 2026 · MIT
Open source ↗
Max output
256Kmodels.devPublished source fact Retrieved Oct 9, 2026 · MIT
Open source ↗
Released
May 29, 2026models.devPublished source fact Retrieved Oct 9, 2026 · MIT
Open source ↗
Open weights
Yesmodels.devPublished source fact Retrieved Oct 9, 2026 · MIT
Open source ↗
License
not yet reported
Input modalities
text, image, videomodels.devPublished source fact Retrieved Oct 9, 2026 · MIT
Open source ↗

Benchmark results

5 results
Benchmark Raw result Normalized (0–100) Pillar
hle tools 47.2%Official model cards via models.devLab-reported; metric accuracy; transcribed by MIT models.dev catalog; not independently evaluated [variant] with toolsPublished May 29, 2026 Retrieved Oct 9, 2026 · factual citation; MIT transcription
Open source ↗ 		47.2 reasoning
swe bench pro 56.3%Official model cards via models.devLab-reported; metric resolve rate; transcribed by MIT models.dev catalog; not independently evaluated [variant] Published May 29, 2026 Retrieved Oct 9, 2026 · factual citation; MIT transcription
Open source ↗ 		56.3 coding
swe bench verified 76.5%Official model cards via models.devLab-reported; metric resolved; transcribed by MIT models.dev catalog; not independently evaluated [variant] Published May 29, 2026 Retrieved Oct 9, 2026 · factual citation; MIT transcription
Open source ↗ 		76.5 coding
terminal bench 35.6%Official model cards via models.devLab-reported; metric success rate; transcribed by MIT models.dev catalog; not independently evaluated [variant] Published Jun 15, 2026 Retrieved Oct 9, 2026 · factual citation; MIT transcription
Open source ↗ 		35.6 coding
terminal bench v2 1 59.6%Official model cards via models.devLab-reported; metric success rate; transcribed by MIT models.dev catalog; not independently evaluated [variant] 2.1Published May 29, 2026 Retrieved Oct 9, 2026 · factual citation; MIT transcription
Open source ↗ 		59.6 coding

Normalization uses fixed absolute 0–100 scales for each unit, independently of other models that have a result on each benchmark. Coverage and evidence breadth still affect the composite; compare the evaluation conditions before reading a small score gap as decisive. “lab-reported” marks the provider's own published figure.

Sources: in and pending

in Reported (1)

  • Official model cards via models.dev · arrived Oct 8, 2026

pending Awaiting (2)

  • LiveBench · carries 30% of expected weight
  • LMArena / Arena · carries 60% of expected weight

The confidence % rises as pending sources publish. Some sources never cover some models — that is why 100% confidence arrives at 80% of expected weight, not at full coverage.