Step 3.7 Flash

Step 3.7 Flash is a reasoning model from StepFun. 22 benchmarks count toward its score, in 6 categories.

availableShows if the model has enough results for an index.
IndexOverall score. 50 is the middle.56.4 ±4.9
CoverageShare of the index weight with results.85%
SpeedOutput tokens per second.93/s
Input / 1MUS dollars per 1M input tokens.$0.2
Output / 1MUS dollars per 1M output tokens.$1.15
ContextMaximum tokens in one request.262K
EloLMArena rating and rank.N/A

50 is the middle of the board. The range shows the doubt in the index.

CapabilitiesScore per category. 50 is the middle.

50 is the middle
AgenticMulti-step tasks with tools.
62.2
CodingCode writing and repair.
53.0
ReasoningLogic problems and puzzles.
51.2
MultimodalTasks with images and text.
62.6
KnowledgeFacts and expert knowledge.
47.9
MultilingualTasks in many languages.
N/A
InstructionTasks with strict rules in the prompt.
57.4
MathMath problems.
N/A

Results

22 counted
BenchmarkThe test name.CategoryThe capability that the test measures.ResultThe score from the publisher.IndexThis result on the index scale.RunThe settings of the run.DateDate of the result.Published byThe source of the result.
τ²-Bench Tool-Agent-User EvaluationAgentic98.5%69.5Victor Barres et al.
V*Multimodal95.3%55.6Z.AI
DeepSearchQAAgentic92.8%71.9Meta AI
Artificial Analysis GPQA DiamondKnowledge80.9%51.5Artificial Analysis
SimpleVQAMultimodal79.2%73.2Z.AI
BrowseCompAgentic75.8%63.5OpenAI
Artificial Analysis MMMU-ProMultimodal75.3%59.1Artificial Analysis
Artificial Analysis Long Context ReasoningReasoning73.7%59.2Artificial Analysis
Artificial Analysis IFBenchInstruction67.3%57.4Artificial Analysis
Claw-EvalAgentic67.1%63.9Bowen Ye et al.
SWE-bench ProCoding56.3%58.4Xiang Deng et al.
Gert Labs Composite Game BenchmarkAgentic51.6%58.3Gert Labs
ToolathlonAgentic49.5%64.1OpenAI
Humanity's Last Exam with toolsAgentic47.2%60.3DeepSeek-AI
Artificial Analysis SciCodeCoding43.9%53.8Artificial Analysis
Artificial Analysis Coding IndexCoding39.6%46.8Artificial Analysis
GDPval-AA normalizedAgentic25.8%55.4Artificial Analysis
Artificial Analysis Omniscience AccuracyKnowledge25.8%45.8Artificial Analysis
Artificial Analysis Humanity's Last ExamKnowledge21.4%47.8Artificial Analysis
Artificial Analysis Intelligence IndexKnowledge19.5%46.8Artificial Analysis
APEX-Agents-AAAgentic14.8%52.8Artificial Analysis / Mercor
Critical Physics TasksReasoning2.3%43.2Artificial Analysis

22 benchmarks count, from 22 of 22 results. A grey row does not count. Too few models took that benchmark.

Sources

BenchLM benchmark aggregationUsed with attribution; per-benchmark results credited to their original authorsOpenRouter, collected directlyNo licence stated

Same level, lower price

Qwen3.8-Flash-Next66.4 · Freedots3-note Preview65.5 · FreeApodex 1.1 Mini59.1 · Free

More from StepFun

Step 5 Preview72.5Step 3.5 FlashUnrankedStep-AudioUnrankedStep-Audio-Chat 130BUnranked