Hugging FaceA 197B, 12B active chat model from StepFun that reads pictures as well as text.
- Downloads
- 17K
- Likes
- 242
- Parameters
- 197B, 12B active
- Context
- 256K
- Licence
- apache-2.0
Too big for this Mac
Its smallest size needs 64 GB of memory, more than M4 · 16 GB can give it.
Ratings
Humanity's Last Exam: questions written to be beyond today's models. Scores are the ones filed on the model's Hugging Face page or the board's own table, read 2026-11-05; some come from the maker.
Step 3.7 Flash runs well from a Mac with an M2 Max and 96 GB of memory. 15 of 59 Mac configurations run it well.
Compare with
The nearest models in size.
- Step 3.5 FlashStepFun · 13K downloads · Arena 1393Too big for this Mac · 119 GB
- Qwen3.8 Flash NextAlibaba Qwen · 9.4M downloadsToo big for this Mac · 66 GB
- Swift 1.5 Qwen3.8 Flash Next GSQ RCOukisai · 771K downloadsToo big for this Mac · 67 GB
- Qwen3.8 Flash Next Gyroagentionai · 749K downloadsToo big for this Mac · 58 GB
Sizes
From stepfun-ai/Step-3.7-Flash-GGUF on Hugging Face. A smaller size is a smaller download and a little less accurate.
| Size | Download | Memory at 8K | Get |
|---|---|---|---|
| IQ3_XXS | 76 GB | 76 GBestimated | Open |
| Q3_K_M | 94 GB | 94 GBestimated | Open |
| Q3_K_L | 103 GB | 103 GBestimated | Open |
| IQ4_XS | 105 GB | 105 GBestimated | Open |
| Q4_K_S | 111 GB | 112 GBestimated | Open |
| Q8_0 | 209 GB | 210 GBestimated | Open |
Run It Anywhere
Step 3.7 Flash is an open model, so any app that runs GGUF files can run it. These take the IQ3_XXS size from stepfun-ai straight from Hugging Face.
- LM StudioMac, Windows, Linux
Open - JanMac, Windows, Linux
Open - MstyMac, Windows, Linux
Open - OllamaMac, Windows, Linux
ollama run hf.co/stepfun-ai/Step-3.7-Flash-GGUF:IQ3_XXS - llama.cppMac, Windows, Linux
llama-server -hf stepfun-ai/Step-3.7-Flash-GGUF:IQ3_XXS - Docker Model RunnerMac, Windows, Linux
docker model run hf.co/stepfun-ai/Step-3.7-Flash-GGUF:IQ3_XXS
On a PC Graphics Card
| Video memory | Largest size that fits | Download | Memory at 8K |
|---|---|---|---|
| 6 GB | None worth running | needs 76 GB | |
| 8 GB | None worth running | needs 76 GB | |
| 10 GB | None worth running | needs 76 GB | |
| 11 GB | None worth running | needs 76 GB | |
| 12 GB | None worth running | needs 76 GB | |
| 16 GB | None worth running | needs 76 GB | |
| 20 GB | None worth running | needs 76 GB | |
| 24 GB | None worth running | needs 76 GB | |
| 32 GB | None worth running | needs 76 GB | |
| 48 GB | None worth running | needs 76 GB |
The largest size, of those that keep three bits a weight or more, whose memory at an 8K chat is under 92% of the card's; the rest is the desktop's. Estimated from the files, the same way as for a Mac; speed on a PC is not estimated yet. A size that doesn't fit the card still runs, slower, with part of it in ordinary memory.
On every Mac
| Mac | Size | Fit | Memory | Writes |
|---|---|---|---|---|
| M6 · 16 GB | UD-IQ1_M | Too big | 64 GB | |
| M6 · 24 GB | UD-IQ1_M | Too big | 64 GB | |
| M6 · 32 GB | UD-IQ1_M | Too big | 64 GB | |
| M5 · 16 GB | UD-IQ1_M | Too big | 64 GB | |
| M5 · 24 GB | UD-IQ1_M | Too big | 64 GB | |
| M5 · 32 GB | UD-IQ1_M | Too big | 64 GB | |
| M5 Pro · 24 GB | UD-IQ1_M | Too big | 64 GB | |
| M5 Pro · 48 GB | UD-IQ1_M | Too big | 64 GB | |
| M5 Pro · 64 GB | UD-IQ1_M | Too big | 64 GB | |
| M5 Max · 36 GB | UD-IQ1_M | Too big | 64 GB | |
| M5 Max · 48 GB | UD-IQ1_M | Too big | 64 GB | |
| M5 Max · 64 GB | UD-IQ1_M | Too big | 64 GB | |
| M5 Max · 128 GB | UD-IQ4_NL | Runs well | 99 GB | 29 |
| M5 Ultra · 96 GB | UD-Q2_K_XL | Runs well | 68 GB | 66 |
| M5 Ultra · 256 GB | UD-Q6_K | Runs great | 171 GB | 37 |
| M5 Ultra · 512 GB | Q8_0 | Runs great | 217 GB | 36 |
| M4 · 16 GB | UD-IQ1_M | Too big | 64 GB | |
| M4 · 24 GB | UD-IQ1_M | Too big | 64 GB | |
| M4 · 32 GB | UD-IQ1_M | Too big | 64 GB | |
| M4 Pro · 24 GB | UD-IQ1_M | Too big | 64 GB | |
| M4 Pro · 48 GB | UD-IQ1_M | Too big | 64 GB | |
| M4 Pro · 64 GB | UD-IQ1_M | Too big | 64 GB | |
| M4 Max · 36 GB | UD-IQ1_M | Too big | 64 GB | |
| M4 Max · 48 GB | UD-IQ1_M | Too big | 64 GB | |
| M4 Max · 64 GB | UD-IQ1_M | Too big | 64 GB | |
| M4 Max · 96 GB | UD-Q2_K_XL | Runs well | 68 GB | 21 |
| M4 Max · 128 GB | UD-IQ4_NL | Runs well | 99 GB | 21 |
| M3 · 8 GB | UD-IQ1_M | Too big | 64 GB | |
| M3 · 16 GB | UD-IQ1_M | Too big | 64 GB | |
| M3 · 24 GB | UD-IQ1_M | Too big | 64 GB | |
| M3 Pro · 18 GB | UD-IQ1_M | Too big | 64 GB | |
| M3 Pro · 36 GB | UD-IQ1_M | Too big | 64 GB | |
| M3 Max · 36 GB | UD-IQ1_M | Too big | 64 GB | |
| M3 Max · 48 GB | UD-IQ1_M | Too big | 64 GB | |
| M3 Max · 64 GB | UD-IQ1_M | Too big | 64 GB | |
| M3 Max · 96 GB | UD-Q2_K_XL | Runs well | 68 GB | 17 |
| M3 Max · 128 GB | UD-IQ4_NL | Runs well | 99 GB | 17 |
| M3 Ultra · 96 GB | UD-Q2_K_XL | Runs well | 68 GB | 34 |
| M3 Ultra · 256 GB | UD-Q6_K | Runs great | 171 GB | 21 |
| M3 Ultra · 512 GB | Q8_0 | Runs great | 217 GB | 20 |
| M2 · 8 GB | UD-IQ1_M | Too big | 64 GB | |
| M2 · 16 GB | UD-IQ1_M | Too big | 64 GB | |
| M2 · 24 GB | UD-IQ1_M | Too big | 64 GB | |
| M2 Pro · 16 GB | UD-IQ1_M | Too big | 64 GB | |
| M2 Pro · 32 GB | UD-IQ1_M | Too big | 64 GB | |
| M2 Max · 32 GB | UD-IQ1_M | Too big | 64 GB | |
| M2 Max · 64 GB | UD-IQ1_M | Too big | 64 GB | |
| M2 Max · 96 GB | UD-Q2_K_XL | Runs well | 68 GB | 21 |
| M2 Ultra · 64 GB | UD-IQ1_M | Too big | 64 GB | |
| M2 Ultra · 128 GB | UD-IQ4_NL | Runs well | 99 GB | 28 |
| M2 Ultra · 192 GB | UD-Q4_K_M | Runs great | 130 GB | 25 |
| M1 · 8 GB | UD-IQ1_M | Too big | 64 GB | |
| M1 · 16 GB | UD-IQ1_M | Too big | 64 GB | |
| M1 Pro · 16 GB | UD-IQ1_M | Too big | 64 GB | |
| M1 Pro · 32 GB | UD-IQ1_M | Too big | 64 GB | |
| M1 Max · 32 GB | UD-IQ1_M | Too big | 64 GB | |
| M1 Max · 64 GB | UD-IQ1_M | Too big | 64 GB | |
| M1 Ultra · 64 GB | UD-IQ1_M | Too big | 64 GB | |
| M1 Ultra · 128 GB | UD-IQ4_NL | Runs well | 99 GB | 23 |
Writes is words a second, estimated from the Mac's memory speed and the model's size; how, and how wrong.
Packaged by
2 repositories on Hugging Face carry this model for llama.cpp. Outmake reads them all as one model.
Information
- Maker
- StepFun
- Released
- May 2026
- Parameters
- 197B, 12B active
- Longest chat
- 256K tokens
- Reads pictures
- Yes
- Licence
- apache-2.0
- Downloads this month
- 16,836
- Likes
- 242
- Sizes
- 6, 76 to 209 GB
- Kind
- Maker's release
Badge for the model card
Shows on Step 3.7 Flash's Hugging Face card which Macs run it well, linked to this page. It updates every morning.
Files, downloads and likes read from Hugging Face on November 5, 2026 and refreshed every morning. Memory and speed are estimates until someone running Outmake measures this model on that Mac.