[whatmodel.sh](../../README.md) / Step 3.7 Flash (198B-A11B) # Step 3.7 Flash (198B-A11B) StepFun · released 2026-05 · 198B params (11B active) · apache-2.0 **score 65.7** ±1.0 · #135 of 732 · from 13 benchmark results across 2 sources Score within the [LMArena (text)](https://whatmodel.sh/?set=lmarena-text&score=set) comparison set only. ## What it's good at ● marks Step 3.7 Flash (198B-A11B) among every model on that task or test, worst to best. 
task         vs the others            rank  counts  tests
agent tasks                 not ranked yet       —      3
coding                      not ranked yet       —      1
math         ▁▁▃▅█▆▄●▄▂▁▁      #104 of 431     93%     11
agent tasks · 0.0% of its score 
test                                result  vs the others in its run  difficulty            counts  why
Terminal-Bench 2.1 (maker)           59.5%  ▁·▁·▁▁▄▃●█▅█ beats 26/63  very hard · avg 7.2%       —  not counted yet: no other model in the …
SWE-bench Pro (split unspecified)…   56.3%  ▂·▁▁▁▁▂▃●▇▂▁ beats 36/58  very hard · avg 3.9%       —  not counted yet: no other model in the …
Toolathlon (Tool Decathlon) (make…   49.5%  ▂▂▄▂▄▄▆▂●▆██ beats 16/27                             —  not counted yet: no other model in the …
coding · 0.0% of its score 
test                                result  vs the others in its run  difficulty            counts  why
SWE-bench Pro (split unspecified)…   56.3%  ▂·▁▁▁▁▂▃●▇▂▁ beats 36/58  very hard · avg 3.9%       —  not counted yet: no other model in the …
math · #104 of 431 · 93% of its score Beats [Claude Sonnet 4.6](claude-sonnet-4-6.md), [MiniMax-M3](minimax-m3.md), [Nemotron 3 Ultra 550B-A55B](nemotron-3-ultra-550b-a55b.md) here, though they rank higher overall. Trails [Gemini 2.5 Flash](gemini-2-5-flash.md), [GLM-4.5-Air](glm-4-5-air.md), [Granite 4.2 30B](granite-4-2-30b.md) here, though they rank lower overall. 
test                                result  vs the others in its run  difficulty            counts  why
MathArena ArXivMath June 2026 (48…   30.6%  ▂···●·██▆▆▄▂ beats 2/19   very hard · avg 5.1%     24%  sharp · at its level
MathArena ArXivMath April 2026 (4…   35.8%  ▁·····●██▄·▁ beats 3/20   very hard · avg 7.6%     21%  sharp · at its level
MathArena ArXivMath February 2026…   32.0%  ▁▁▁▁●▇▄▃▁▃·▃ beats 7/24   very hard · avg 9.4%     12%  at its level
MathArena ArXivMath March 2026 (3…   45.0%  ●▃▅·█▃▃·▅·▃▃ beats 1/14   hard · avg 16%           11%  at its level
MathArena ArXivMath May 2026 (40 …   46.7%  ▁··▁●█▄▁·▃·▁ beats 3/17   very hard · avg 11%     8.0%  at its level
MathArena BrokenArXiv April 2026 …   18.9%  ▂▂●█▃▃▂▂▂··▂ beats 5/18   very hard · avg 0.8%    3.8%  sharp · noisy
MathArena BrokenArXiv June 2026 (…   15.7%  ▄··●▂▂▂▆▄··▂ beats 5/14   very hard · avg 1.3%    3.0%  sharp · noisy
MathArena BrokenArXiv May 2026 (5…    8.5%  ●█·▄▄▄····▄▄ beats 0/8    very hard · avg 1.9%    3.0%  sharp
MathArena ArXivLean June 2026 (46…    0.0%  ●····▄▄███·▄ beats 0/9    very hard · avg 0.9%    2.7%  sharp · too hard for it
MathArena BrokenArXiv March 2026 …   10.7%  █·●█▂·▂▂···▂ beats 4/14   very hard · avg 1.0%    2.2%  sharp · noisy
MathArena BrokenArXiv February 20…   12.1%  ▃▂▂▃●▃▂·▃··▂ beats 9/16   very hard · avg 0.9%    2.1%  sharp · too hard for it · noisy
other tests (5) 
test                                result  vs the others in its run  difficulty            counts  why
WideSearch corrected dataset         56.8%  ▃······●▃▃█▃ beats 1/7    hard · avg 22%          3.4%  at its level · noisy
SkillsBench 87-task subset           24.9%  ▄··▄●▄···█·█ beats 2/7    very hard · avg 9.1%    3.1%  at its level
SimpleVQA with tools (search) (ma…   79.2%  ▄·······█··● beats 3/3    hard · avg 27%             —  not counted yet: no other model in the …
V* Bench with Python tool (maker)    95.3%  no other model yet                                   —  not counted yet: no other model in the …
ClawEval 1.1 (metric unspecified)…   67.1%  no other model yet                                   —  not counted yet: no other model in the …
## How the score was worked out Each result implies a score on its own, after adjusting for how hard that test is and who ran it. The model's score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its *counts* column above; here it is added up by who ran the tests. [Method](../method.md) 
source                                    results  share of the score
MathArena ArXivMath / BrokenArXiv / ArX…       11   93% ▇▇▇▇▇▇▇▇▇▇▇
inclusionAI/Ling-3.0-flash model card           2    6% ▇
prior: pull toward the middle                        0% ▇
weighted average                                   = 65.7
Licences: MathArena ArXivMath / BrokenArXiv / ArXivLean model-output datasets ([CC BY-SA 4.0](../../licenses/matharena-sharealike-NOTICE.txt)) · inclusionAI/Ling-3.0-flash model card ([MIT](../../licenses/cards-final-NOTICE.txt))
Not counted (6) 
source                  test                            score  why
stepfun-ai/Step-3.7-F…  SWE-bench Pro (split unspecif…  56.3%  too few models have it yet
stepfun-ai/Step-3.7-F…  Terminal-Bench 2.1              59.5%  too few models have it yet
stepfun-ai/Step-3.7-F…  Toolathlon (Tool Decathlon)     49.5%  too few models have it yet
stepfun-ai/Step-3.7-F…  SimpleVQA with tools (search)   79.2%  too few models have it yet
stepfun-ai/Step-3.7-F…  V* Bench with Python tool       95.3%  too few models have it yet
stepfun-ai/Step-3.7-F…  ClawEval 1.1 (metric unspecif…  67.1%  too few models have it yet
## How to run it ### Through an API api (StepFun) · $0.44/M blended · $0.2 in / $1.15 out per 1M tokens · [platform.stepfun.ai](https://platform.stepfun.ai/docs/en/guides/models/step-3.7-flash.md) [price source](https://platform.stepfun.ai/docs/en/guides/pricing/details.md): USD per 1M tokens, StepFun Open Platform (international), step-3.7-flash (Multimodal Reasoning Models table): input on cache miss $0.20, output $1.15 (output includes reasoning). Cache-hit input ($0.04/1M) excluded. Currency: USD. StepFun official pricing page, observed 2026-09-26. · [context source](https://platform.stepfun.ai/docs/en/guides/models/step-3.7-flash.md): StepFun official model page, step-3.7-flash: 'Context length 256K tokens' (a rounded figure; models.dev @a668908 gives 256,000). Observed 2026-09-26. ### llama.cpp (recommended) 1. Install llama.cpp. Homebrew and winget installs name the commands llama-cli and llama-server instead. [Other ways to install](https://github.com/ggml-org/llama.cpp/blob/master/docs/install.md) ```sh # macOS / Linux curl -LsSf https://llama.app/install.sh | sh # Windows irm https://llama.app/install.ps1 | iex ``` 2. Run Step 3.7 Flash (198B-A11B) (Q4_K_S, needs ~115.47 GB, from [huggingface.co/stepfun-ai](https://huggingface.co/stepfun-ai/Step-3.7-Flash-GGUF/tree/0b69336d2fd2adfdef9c66e425f7778196c31482)) ```sh llama cli -hf stepfun-ai/Step-3.7-Flash-GGUF:Q4_K_S ``` 3. Or start a local server and open the address it prints (http://localhost:9931) ```sh llama serve -hf stepfun-ai/Step-3.7-Flash-GGUF:Q4_K_S ``` Other versions: Q3_K_M (~97.77 GB) `llama cli -hf stepfun-ai/Step-3.7-Flash-GGUF:Q3_K_M` ### LM Studio 1. Install LM Studio. Open the app once before using the lms command. [Download it (macOS / Windows / Linux)](https://lmstudio.ai/download) 2. Run Step 3.7 Flash (198B-A11B) (Q4_K_S, needs ~115.47 GB, from [huggingface.co/stepfun-ai](https://huggingface.co/stepfun-ai/Step-3.7-Flash-GGUF/tree/0b69336d2fd2adfdef9c66e425f7778196c31482)) ```sh lms get https://huggingface.co/stepfun-ai/Step-3.7-Flash-GGUF@Q4_K_S -y ``` 3. Then run lms ls to see its name, and lms chat followed by that name Other versions: Q3_K_M (~97.77 GB) `lms get https://huggingface.co/stepfun-ai/Step-3.7-Flash-GGUF@Q3_K_M -y` Model repository and documentation: [huggingface.co/stepfun-ai](https://huggingface.co/stepfun-ai/Step-3.7-Flash) [interactive version](https://whatmodel.sh/model/step-3-7-flash) · [JSON](https://whatmodel.sh/api?format=json&all=1)