The short version
- The best coders solved 29 of 30 jobs: Qwen3.6 27B.
- The jobs are things like reading dates, cleaning up messy data and building a simple stock tracker.
- Below: the best coder that fits each graphics card and Mac.
Pick by graphics card
Pick by Mac
All models, ranked by coding
| # | Model | Coding | Runs on (PC) | Runs on (Mac) |
|---|---|---|---|---|
| 1 | Qwen3.6 27B 28B | 97 | 24 GB card (Q5) | 32 GB Mac (Q4) |
| 2 | Gemma 4 31B 31B | 93 | 32 GB card (Q5) | 48 GB Mac (Q6) |
| 3 | gpt-oss-120b 117B · gpt-oss:120b | 93 | more than 32 GB | 96 GB Mac (standard) |
| 4 | Ling 3.0 Flash 127B | 93 | more than 32 GB | 128 GB Mac (Q4) |
| 5 | gpt-oss-20b 21B · gpt-oss:20b | 90 | 16 GB card (standard) | 24 GB Mac (standard) |
| 6 | Muse Glimmer 30B 30B | 90 | 24 GB card (Q5) | 32 GB Mac (Q4) |
| 7 | GLM 4.5V 108B | 87 | more than 32 GB | 96 GB Mac (Q4) |
| 8 | Ling 3.0 Flash VL 125B | 87 | more than 32 GB | 128 GB Mac (Q5) |
| 9 | Gemma 4 26B A4B 26B | 83 | 24 GB card (Q5) | 32 GB Mac (Q5) |
| 10 | Qwen3.5-35B-A3B 36B | 80 | 24 GB card (Q4) | 48 GB Mac (Q6) |
| 11 | Qwen3.6 35B A3B 36B | 80 | 24 GB card (Q4) | 48 GB Mac (Q6) |
| 12 | Granite 4.2 8B 8.8B | 77 | 8 GB card (Q4) | 16 GB Mac (Q6) |
| 13 | Qwen3.5-27B 28B | 77 | 24 GB card (Q5) | 32 GB Mac (Q4) |
| 14 | Qwen3.8 27B 28B | 77 | 24 GB card (Q5) | 32 GB Mac (Q4) |
| 15 | GLM 4.7 Flash 31B | 73 | 24 GB card (Q4) | 48 GB Mac (Q6) |
| 16 | Laguna XS 2.1 33B | 73 | 24 GB card (Q4) | 48 GB Mac (Q6) |
| 17 | GLM 4.6V 108B | 73 | more than 32 GB | 96 GB Mac (Q4) |
| 18 | Nemotron 3 Nano 30B A3B 32B | 70 | 24 GB card (Q5) | 32 GB Mac (Q4) |
| 19 | GLM 4.5 Air 110B | 70 | more than 32 GB | 96 GB Mac (Q4) |
| 20 | Laguna S 2.1 118B | 70 | more than 32 GB | 128 GB Mac (Q5) |
| 21 | Qwen3 30B A3B 31B · qwen3:30b | 67 | 24 GB card (Q5) | 32 GB Mac (Q4) |
| 22 | Qwen3 14B 15B · qwen3:14b | 63 | 12 GB card (Q4) | 24 GB Mac (Q6) |
| 23 | Qwen3 VL 30B A3B Thinking 31B | 60 | 24 GB card (Q5) | 32 GB Mac (Q4) |
| 24 | Qwen3 Coder 30B A3B Instruct 31B · qwen3-coder:30b | 57 | 24 GB card (Q5) | 32 GB Mac (Q4) |
| 25 | Devstral 2 2512 125B | 53 | more than 32 GB | 128 GB Mac (Q5) |
| 26 | Qwen3 32B 33B · qwen3:32b | 50 | 24 GB card (Q4) | 48 GB Mac (Q6) |
| 27 | Qwen3.5-9B 9.7B | 47 | 8 GB card (Q4) | 16 GB Mac (Q6) |
| 28 | Llama 3.3 70B Instruct 71B | 47 | more than 32 GB | 64 GB Mac (Q4) |
| 29 | Ministral 3 14B 2512 14B | 43 | 12 GB card (Q4) | 16 GB Mac (Q4) |
| 30 | Qwen3 VL 30B A3B Instruct 31B | 43 | 24 GB card (Q5) | 32 GB Mac (Q4) |
| 31 | Nemotron 3.5 Lightning 32B | 43 | 24 GB card (Q5) | 32 GB Mac (Q4) |
| 32 | Mistral Small 3.2 24B 24B | 40 | 24 GB card (Q6) | 32 GB Mac (Q5) |
| 33 | Qwen2.5 72B Instruct 73B | 40 | more than 32 GB | 64 GB Mac (Q4) |
| 34 | Qwen3 VL 8B Instruct 8.8B | 37 | 8 GB card (Q4) | 16 GB Mac (Q6) |
| 35 | Mistral Small 3 24B | 37 | 24 GB card (Q6) | 32 GB Mac (Q5) |
| 36 | Hunyuan A13B Instruct 80B | 37 | more than 32 GB | 96 GB Mac (Q6) |
| 37 | Gemma 3 27B 27B · gemma3:27b | 33 | 24 GB card (Q5) | 32 GB Mac (Q4) |
| 38 | Qwen3 30B A3B Instruct 2507 31B | 33 | 24 GB card (Q5) | 32 GB Mac (Q4) |
| 39 | Llama 3.1 70B Instruct 71B | 33 | more than 32 GB | 64 GB Mac (Q4) |
| 40 | Command A 111B | 33 | more than 32 GB | 96 GB Mac (Q4) |
| 41 | Ministral 3 8B 2512 8.9B | 30 | 8 GB card (Q4) | 16 GB Mac (Q6) |
| 42 | Gemma 3 12B 12B · gemma3:12b | 30 | 12 GB card (Q5) | 16 GB Mac (Q5) |
| 43 | Nex-N2.5-Mini 35B | 27 | 24 GB card (Q4) | 48 GB Mac (Q6) |
| 44 | Phi 4 15B · phi4 | 23 | 12 GB card (Q4) | 24 GB Mac (Q6) |
| 45 | Granite 4.0 Micro (Q4, Mac) 3.2B · granite4:micro-h | 13 | 8 GB card (Q4) | 16 GB Mac (Q4) |
| 46 | Llama 3.3 8B Instruct (Q4, Mac) 8B · hf.co/bartowski/allura-forge_Llama-3.3-8B-Instruct-GGUF:Q4_K_M | 13 | 8 GB card (Q4) | 16 GB Mac (Q4) |
| 47 | Gemma 2 27B 27B · gemma2:27b | 13 | 24 GB card (Q5) | 32 GB Mac (Q4) |
| 48 | Qwen2.5 VL 72B Instruct 73B | 13 | more than 32 GB | 64 GB Mac (Q4) |
| 49 | Ministral 3 3B 2512 3.9B | 10 | 8 GB card (Q8) | 16 GB Mac (Q8) |
| 50 | Qwen2.5 7B Instruct 7.6B · qwen2.5:7b | 10 | 8 GB card (Q6) | 16 GB Mac (Q8) |
| 51 | Mistral Nemo 12B · mistral-nemo | 10 | 12 GB card (Q5) | 16 GB Mac (Q4) |
| 52 | Gemma 3 4B 4.3B · gemma3:4b | 3 | 8 GB card (Q8) | 16 GB Mac (Q8) |
| 53 | Llama 3.1 8B Instruct 8B · llama3.1:8b | 3 | 8 GB card (Q5) | 16 GB Mac (Q8) |
| 54 | Llama 3.2 1B Instruct (Q4, Mac) 1.2B · llama3.2:1b-instruct-q4_K_M | 0 | 8 GB card (Q4) | 16 GB Mac (Q4) |
| 55 | Llama 3.2 3B Instruct 3.2B · llama3.2:3b | 0 | 8 GB card (Q8) | 16 GB Mac (Q8) |
| 56 | Reka Edge 7.1B | 0 | 8 GB card (Q6) | 16 GB Mac (Q8) |
Scores are out of 100. "Runs on" is the smallest graphics card or Mac that can hold the model, and which version to download for it. How we worked this out
What this test doesn't cover
These are short jobs that fit in one file, the kind of thing you'd ask a chatbot for. We don't test big projects spread over many files, and we don't measure how fast a model runs on your machine. You can see five of the jobs, and every model's answer, on each model's review page. Full technical details.
Comments
Sign in with GitHub to comment. Spam and abuse are hidden automatically.