Best local LLM for translation · 2026
The best local LLM for translation, ranked for your VRAM
The translation shortlist is ordered by language coverage and by how translations read in community blind comparisons, with room for specialists such as LLM-jp for Japanese, as of the catalogue date. For each memory class below we list every model on that list that actually fits at Q4_K_M with an 8K context, so the top row is the pick and the rest are what you give up by choosing it. Numbers are computed, not benchmarked; how the shortlist works.
Coverage is not quality. A model pretrained on 140 languages handles the large European and East Asian ones well and the long tail unevenly — for a low-resource language, test it on a paragraph you can check before you trust it with a whole document.
Best translation LLM for 6 GB VRAM
5.3 GB usable · speeds for a GeForce RTX 3060 Laptop everything that fits 6 GB →Gemma 4 E4B is the pick: 140+ languages on a laptop. 2 other models on the translation shortlist also fit, in order of capability:
| # | Model | Why | Params | Quant · total | Max context | Tok/s est. | |
|---|---|---|---|---|---|---|---|
| 1 | Gemma 4 E4B Google · Apache 2.0 |
140+ languages on a laptop. | 8.0B | Q4_K_M · 5.2 GB | 12K | 45 | Tight fit |
| 2 | Qwen3.5 4B Alibaba · Apache 2.0 |
Usable translation in 8 GB. | 4.66B | Q4_K_M · 3.5 GB | 66K | 78 | Runs great |
| 3 | Qwen3.5 2B Alibaba · Apache 2.0 |
Phrase-level translation on 4 GB. | 2.27B | Q4_K_M · 2.0 GB | 256K | 159 | Runs great |
Best translation LLM for 8 GB VRAM
7.0 GB usable · speeds for a GeForce RTX 4060 everything that fits 8 GB →Qwen3.5 9B is the pick: Strong CJK and European translation in 8–12 GB. 3 other models on the translation shortlist also fit, in order of capability:
| # | Model | Why | Params | Quant · total | Max context | Tok/s est. | |
|---|---|---|---|---|---|---|---|
| 1 | Qwen3.5 9B Alibaba · Apache 2.0 |
Strong CJK and European translation in 8–12 GB. | 9.65B | Q4_K_M · 6.3 GB | 31K | 30 | Tight fit |
| 2 | Gemma 4 E4B Google · Apache 2.0 |
140+ languages on a laptop. | 8.0B | Q4_K_M · 5.2 GB | 128K | 37 | Runs great |
| 3 | Qwen3.5 4B Alibaba · Apache 2.0 |
Usable translation in 8 GB. | 4.66B | Q4_K_M · 3.5 GB | 120K | 63 | Runs great |
| 4 | Qwen3.5 2B Alibaba · Apache 2.0 |
Phrase-level translation on 4 GB. | 2.27B | Q4_K_M · 2.0 GB | 256K | 129 | Runs great |
Best translation LLM for 12 GB VRAM
10.6 GB usable · speeds for a GeForce RTX 3060 12 GB everything that fits 12 GB →Gemma 4 12B is the pick: Broad language coverage on a 12 GB card. 6 other models on the translation shortlist also fit, in order of capability:
| # | Model | Why | Params | Quant · total | Max context | Tok/s est. | |
|---|---|---|---|---|---|---|---|
| 1 | Gemma 4 12B Google · Apache 2.0 |
Broad language coverage on a 12 GB card. | 12B | Q4_K_M · 8.2 GB | 47K | 32 | Runs great |
| 2 | Qwen3.5 9B Alibaba · Apache 2.0 |
Strong CJK and European translation in 8–12 GB. | 9.65B | Q4_K_M · 6.3 GB | 146K | 40 | Runs great |
| 3 | Qwen3 14B Alibaba · Apache 2.0 |
Good European and CJK coverage at 14B. | 14.8B | Q4_K_M · 10.2 GB | 10K | 26 | Tight fit |
| 4 | Gemma 4 E4B Google · Apache 2.0 |
140+ languages on a laptop. | 8.0B | Q4_K_M · 5.2 GB | 128K | 48 | Runs great |
| 5 | Mistral NeMo 12B Mistral AI · Apache 2.0 |
Built multilingual, with a Tekken tokenizer efficient for European languages. | 12.2B | Q4_K_M · 8.7 GB | 20K | 32 | Runs great |
| 6 | Qwen3.5 4B Alibaba · Apache 2.0 |
Usable translation in 8 GB. | 4.66B | Q4_K_M · 3.5 GB | 236K | 83 | Runs great |
| 7 | Qwen3.5 2B Alibaba · Apache 2.0 |
Phrase-level translation on 4 GB. | 2.27B | Q4_K_M · 2.0 GB | 256K | 171 | Runs great |
Best translation LLM for 16 GB VRAM
14.4 GB usable · speeds for a GeForce RTX 5080 everything that fits 16 GB →Gemma 4 12B is the pick: Broad language coverage on a 12 GB card. 6 other models on the translation shortlist also fit, in order of capability:
| # | Model | Why | Params | Quant · total | Max context | Tok/s est. | |
|---|---|---|---|---|---|---|---|
| 1 | Gemma 4 12B Google · Apache 2.0 |
Broad language coverage on a 12 GB card. | 12B | Q4_K_M · 8.2 GB | 107K | 86 | Runs great |
| 2 | Qwen3.5 9B Alibaba · Apache 2.0 |
Strong CJK and European translation in 8–12 GB. | 9.65B | Q4_K_M · 6.3 GB | 256K | 107 | Runs great |
| 3 | Qwen3 14B Alibaba · Apache 2.0 |
Good European and CJK coverage at 14B. | 14.8B | Q4_K_M · 10.2 GB | 35K | 70 | Runs great |
| 4 | Gemma 4 E4B Google · Apache 2.0 |
140+ languages on a laptop. | 8.0B | Q4_K_M · 5.2 GB | 128K | 129 | Runs great |
| 5 | Mistral NeMo 12B Mistral AI · Apache 2.0 |
Built multilingual, with a Tekken tokenizer efficient for European languages. | 12.2B | Q4_K_M · 8.7 GB | 44K | 85 | Runs great |
| 6 | Qwen3.5 4B Alibaba · Apache 2.0 |
Usable translation in 8 GB. | 4.66B | Q4_K_M · 3.5 GB | 256K | 222 | Runs great |
| 7 | Qwen3.5 2B Alibaba · Apache 2.0 |
Phrase-level translation on 4 GB. | 2.27B | Q4_K_M · 2.0 GB | 256K | 455 | Runs great |
Best translation LLM for 24 GB VRAM
22.4 GB usable · speeds for a GeForce RTX 4090 everything that fits 24 GB →Qwen3.8 27B is the pick: 119 languages, and Qwen translations read as idiomatic rather than literal. 11 other models on the translation shortlist also fit, in order of capability:
| # | Model | Why | Params | Quant · total | Max context | Tok/s est. | |
|---|---|---|---|---|---|---|---|
| 1 | Qwen3.8 27B Alibaba · Apache 2.0 |
119 languages, and Qwen translations read as idiomatic rather than literal. | 27.8B | Q4_K_M · 16.7 GB | 98K | 39 | Runs great |
| 2 | Gemma 4 31B Google · Apache 2.0 |
140+ language pretraining — the pick for low-resource languages. | 31.3B | Q4_K_M · 20.2 GB | 21K | 35 | Tight fit |
| 3 | Qwen3 32B Alibaba · Apache 2.0 |
Still the most capable local translator under 70B in blind tests. | 32.8B | Q4_K_M · 21.0 GB | 13K | 33 | Tight fit |
| 4 | LLM-jp 4 33B Thinking NII (Japan) · Apache 2.0 |
Japan’s national-institute model — the pick for Japanese ↔ English. | 33.2B | Q4_K_M · 21.3 GB | 12K | 33 | Tight fit |
| 5 | Gemma 4 26B-A4B Google · Apache 2.0 |
Gemma 4 language coverage at MoE speed. | 26.5B MoE | Q4_K_M · 16.0 GB | 171K | 110 | Runs great |
| 6 | Gemma 4 12B Google · Apache 2.0 |
Broad language coverage on a 12 GB card. | 12B | Q4_K_M · 8.2 GB | 235K | 90 | Runs great |
| 7 | Qwen3.5 9B Alibaba · Apache 2.0 |
Strong CJK and European translation in 8–12 GB. | 9.65B | Q4_K_M · 6.3 GB | 256K | 112 | Runs great |
| 8 | Qwen3 14B Alibaba · Apache 2.0 |
Good European and CJK coverage at 14B. | 14.8B | Q4_K_M · 10.2 GB | 86K | 73 | Runs great |
| 9 | Gemma 4 E4B Google · Apache 2.0 |
140+ languages on a laptop. | 8.0B | Q4_K_M · 5.2 GB | 128K | 136 | Runs great |
| 10 | Mistral NeMo 12B Mistral AI · Apache 2.0 |
Built multilingual, with a Tekken tokenizer efficient for European languages. | 12.2B | Q4_K_M · 8.7 GB | 95K | 89 | Runs great |
| 11 | Qwen3.5 4B Alibaba · Apache 2.0 |
Usable translation in 8 GB. | 4.66B | Q4_K_M · 3.5 GB | 256K | 233 | Runs great |
| 12 | Qwen3.5 2B Alibaba · Apache 2.0 |
Phrase-level translation on 4 GB. | 2.27B | Q4_K_M · 2.0 GB | 256K | 478 | Runs great |
Best translation LLM for 32 GB VRAM
30.4 GB usable · speeds for a GeForce RTX 5090 everything that fits 32 GB →Qwen3.8 27B is the pick: 119 languages, and Qwen translations read as idiomatic rather than literal. 11 other models on the translation shortlist also fit, in order of capability:
| # | Model | Why | Params | Quant · total | Max context | Tok/s est. | |
|---|---|---|---|---|---|---|---|
| 1 | Qwen3.8 27B Alibaba · Apache 2.0 |
119 languages, and Qwen translations read as idiomatic rather than literal. | 27.8B | Q4_K_M · 16.7 GB | 226K | 69 | Runs great |
| 2 | Gemma 4 31B Google · Apache 2.0 |
140+ language pretraining — the pick for low-resource languages. | 31.3B | Q4_K_M · 20.2 GB | 73K | 62 | Runs great |
| 3 | Qwen3 32B Alibaba · Apache 2.0 |
Still the most capable local translator under 70B in blind tests. | 32.8B | Q4_K_M · 21.0 GB | 45K | 59 | Runs great |
| 4 | LLM-jp 4 33B Thinking NII (Japan) · Apache 2.0 |
Japan’s national-institute model — the pick for Japanese ↔ English. | 33.2B | Q4_K_M · 21.3 GB | 44K | 58 | Runs great |
| 5 | Gemma 4 26B-A4B Google · Apache 2.0 |
Gemma 4 language coverage at MoE speed. | 26.5B MoE | Q4_K_M · 16.0 GB | 256K | 195 | Runs great |
| 6 | Gemma 4 12B Google · Apache 2.0 |
Broad language coverage on a 12 GB card. | 12B | Q4_K_M · 8.2 GB | 256K | 161 | Runs great |
| 7 | Qwen3.5 9B Alibaba · Apache 2.0 |
Strong CJK and European translation in 8–12 GB. | 9.65B | Q4_K_M · 6.3 GB | 256K | 200 | Runs great |
| 8 | Qwen3 14B Alibaba · Apache 2.0 |
Good European and CJK coverage at 14B. | 14.8B | Q4_K_M · 10.2 GB | 128K | 130 | Runs great |
| 9 | Gemma 4 E4B Google · Apache 2.0 |
140+ languages on a laptop. | 8.0B | Q4_K_M · 5.2 GB | 128K | 241 | Runs great |
| 10 | Mistral NeMo 12B Mistral AI · Apache 2.0 |
Built multilingual, with a Tekken tokenizer efficient for European languages. | 12.2B | Q4_K_M · 8.7 GB | 128K | 158 | Runs great |
| 11 | Qwen3.5 4B Alibaba · Apache 2.0 |
Usable translation in 8 GB. | 4.66B | Q4_K_M · 3.5 GB | 256K | 414 | Runs great |
| 12 | Qwen3.5 2B Alibaba · Apache 2.0 |
Phrase-level translation on 4 GB. | 2.27B | Q4_K_M · 2.0 GB | 256K | 850 | Runs great |
Best translation LLM for 48 GB VRAM
46.4 GB usable · speeds for an RTX 6000 Ada everything that fits 48 GB →Qwen3.8 27B is the pick: 119 languages, and Qwen translations read as idiomatic rather than literal. 11 other models on the translation shortlist also fit, in order of capability:
| # | Model | Why | Params | Quant · total | Max context | Tok/s est. | |
|---|---|---|---|---|---|---|---|
| 1 | Qwen3.8 27B Alibaba · Apache 2.0 |
119 languages, and Qwen translations read as idiomatic rather than literal. | 27.8B | Q4_K_M · 16.7 GB | 256K | 37 | Runs great |
| 2 | Gemma 4 31B Google · Apache 2.0 |
140+ language pretraining — the pick for low-resource languages. | 31.3B | Q4_K_M · 20.2 GB | 175K | 33 | Runs great |
| 3 | Qwen3 32B Alibaba · Apache 2.0 |
Still the most capable local translator under 70B in blind tests. | 32.8B | Q4_K_M · 21.0 GB | 109K | 32 | Runs great |
| 4 | LLM-jp 4 33B Thinking NII (Japan) · Apache 2.0 |
Japan’s national-institute model — the pick for Japanese ↔ English. | 33.2B | Q4_K_M · 21.3 GB | 64K | 31 | Runs great |
| 5 | Gemma 4 26B-A4B Google · Apache 2.0 |
Gemma 4 language coverage at MoE speed. | 26.5B MoE | Q4_K_M · 16.0 GB | 256K | 105 | Runs great |
| 6 | Gemma 4 12B Google · Apache 2.0 |
Broad language coverage on a 12 GB card. | 12B | Q4_K_M · 8.2 GB | 256K | 86 | Runs great |
| 7 | Qwen3.5 9B Alibaba · Apache 2.0 |
Strong CJK and European translation in 8–12 GB. | 9.65B | Q4_K_M · 6.3 GB | 256K | 107 | Runs great |
| 8 | Qwen3 14B Alibaba · Apache 2.0 |
Good European and CJK coverage at 14B. | 14.8B | Q4_K_M · 10.2 GB | 128K | 70 | Runs great |
| 9 | Gemma 4 E4B Google · Apache 2.0 |
140+ languages on a laptop. | 8.0B | Q4_K_M · 5.2 GB | 128K | 129 | Runs great |
| 10 | Mistral NeMo 12B Mistral AI · Apache 2.0 |
Built multilingual, with a Tekken tokenizer efficient for European languages. | 12.2B | Q4_K_M · 8.7 GB | 128K | 85 | Runs great |
| 11 | Qwen3.5 4B Alibaba · Apache 2.0 |
Usable translation in 8 GB. | 4.66B | Q4_K_M · 3.5 GB | 256K | 222 | Runs great |
| 12 | Qwen3.5 2B Alibaba · Apache 2.0 |
Phrase-level translation on 4 GB. | 2.27B | Q4_K_M · 2.0 GB | 256K | 455 | Runs great |
Best translation LLM for 16 GB unified memory
10.7 GB usable · speeds for an M4 · 16 GB everything that fits 16 GB →Gemma 4 12B is the pick: Broad language coverage on a 12 GB card. 6 other models on the translation shortlist also fit, in order of capability:
| # | Model | Why | Params | Quant · total | Max context | Tok/s est. | |
|---|---|---|---|---|---|---|---|
| 1 | Gemma 4 12B Google · Apache 2.0 |
Broad language coverage on a 12 GB card. | 12B | Q4_K_M · 8.2 GB | 48K | 9.9 | Runs great |
| 2 | Qwen3.5 9B Alibaba · Apache 2.0 |
Strong CJK and European translation in 8–12 GB. | 9.65B | Q4_K_M · 6.3 GB | 149K | 12 | Runs great |
| 3 | Qwen3 14B Alibaba · Apache 2.0 |
Good European and CJK coverage at 14B. | 14.8B | Q4_K_M · 10.2 GB | 11K | 8.1 | Tight fit |
| 4 | Gemma 4 E4B Google · Apache 2.0 |
140+ languages on a laptop. | 8.0B | Q4_K_M · 5.2 GB | 128K | 15 | Runs great |
| 5 | Mistral NeMo 12B Mistral AI · Apache 2.0 |
Built multilingual, with a Tekken tokenizer efficient for European languages. | 12.2B | Q4_K_M · 8.7 GB | 20K | 9.8 | Runs great |
| 6 | Qwen3.5 4B Alibaba · Apache 2.0 |
Usable translation in 8 GB. | 4.66B | Q4_K_M · 3.5 GB | 239K | 26 | Runs great |
| 7 | Qwen3.5 2B Alibaba · Apache 2.0 |
Phrase-level translation on 4 GB. | 2.27B | Q4_K_M · 2.0 GB | 256K | 53 | Runs great |
Best translation LLM for 64 GB unified memory
48.0 GB usable · speeds for an M4 Max · 64 GB everything that fits 64 GB →Qwen3.8 27B is the pick: 119 languages, and Qwen translations read as idiomatic rather than literal. 11 other models on the translation shortlist also fit, in order of capability:
| # | Model | Why | Params | Quant · total | Max context | Tok/s est. | |
|---|---|---|---|---|---|---|---|
| 1 | Qwen3.8 27B Alibaba · Apache 2.0 |
119 languages, and Qwen translations read as idiomatic rather than literal. | 27.8B | Q4_K_M · 16.7 GB | 256K | 20 | Runs great |
| 2 | Gemma 4 31B Google · Apache 2.0 |
140+ language pretraining — the pick for low-resource languages. | 31.3B | Q4_K_M · 20.2 GB | 185K | 17 | Runs great |
| 3 | Qwen3 32B Alibaba · Apache 2.0 |
Still the most capable local translator under 70B in blind tests. | 32.8B | Q4_K_M · 21.0 GB | 115K | 17 | Runs great |
| 4 | LLM-jp 4 33B Thinking NII (Japan) · Apache 2.0 |
Japan’s national-institute model — the pick for Japanese ↔ English. | 33.2B | Q4_K_M · 21.3 GB | 64K | 16 | Runs great |
| 5 | Gemma 4 26B-A4B Google · Apache 2.0 |
Gemma 4 language coverage at MoE speed. | 26.5B MoE | Q4_K_M · 16.0 GB | 256K | 67 | Runs great |
| 6 | Gemma 4 12B Google · Apache 2.0 |
Broad language coverage on a 12 GB card. | 12B | Q4_K_M · 8.2 GB | 256K | 45 | Runs great |
| 7 | Qwen3.5 9B Alibaba · Apache 2.0 |
Strong CJK and European translation in 8–12 GB. | 9.65B | Q4_K_M · 6.3 GB | 256K | 56 | Runs great |
| 8 | Qwen3 14B Alibaba · Apache 2.0 |
Good European and CJK coverage at 14B. | 14.8B | Q4_K_M · 10.2 GB | 128K | 37 | Runs great |
| 9 | Gemma 4 E4B Google · Apache 2.0 |
140+ languages on a laptop. | 8.0B | Q4_K_M · 5.2 GB | 128K | 68 | Runs great |
| 10 | Mistral NeMo 12B Mistral AI · Apache 2.0 |
Built multilingual, with a Tekken tokenizer efficient for European languages. | 12.2B | Q4_K_M · 8.7 GB | 128K | 44 | Runs great |
| 11 | Qwen3.5 4B Alibaba · Apache 2.0 |
Usable translation in 8 GB. | 4.66B | Q4_K_M · 3.5 GB | 256K | 116 | Runs great |
| 12 | Qwen3.5 2B Alibaba · Apache 2.0 |
Phrase-level translation on 4 GB. | 2.27B | Q4_K_M · 2.0 GB | 256K | 239 | Runs great |
Best translation LLM for 128 GB unified memory
96.0 GB usable · speeds for an M4 Max · 128 GB everything that fits 128 GB →Qwen3.8 27B is the pick: 119 languages, and Qwen translations read as idiomatic rather than literal. 11 other models on the translation shortlist also fit, in order of capability:
| # | Model | Why | Params | Quant · total | Max context | Tok/s est. | |
|---|---|---|---|---|---|---|---|
| 1 | Qwen3.8 27B Alibaba · Apache 2.0 |
119 languages, and Qwen translations read as idiomatic rather than literal. | 27.8B | Q4_K_M · 16.7 GB | 256K | 20 | Runs great |
| 2 | Gemma 4 31B Google · Apache 2.0 |
140+ language pretraining — the pick for low-resource languages. | 31.3B | Q4_K_M · 20.2 GB | 256K | 17 | Runs great |
| 3 | Qwen3 32B Alibaba · Apache 2.0 |
Still the most capable local translator under 70B in blind tests. | 32.8B | Q4_K_M · 21.0 GB | 128K | 17 | Runs great |
| 4 | LLM-jp 4 33B Thinking NII (Japan) · Apache 2.0 |
Japan’s national-institute model — the pick for Japanese ↔ English. | 33.2B | Q4_K_M · 21.3 GB | 64K | 16 | Runs great |
| 5 | Gemma 4 26B-A4B Google · Apache 2.0 |
Gemma 4 language coverage at MoE speed. | 26.5B MoE | Q4_K_M · 16.0 GB | 256K | 67 | Runs great |
| 6 | Gemma 4 12B Google · Apache 2.0 |
Broad language coverage on a 12 GB card. | 12B | Q4_K_M · 8.2 GB | 256K | 45 | Runs great |
| 7 | Qwen3.5 9B Alibaba · Apache 2.0 |
Strong CJK and European translation in 8–12 GB. | 9.65B | Q4_K_M · 6.3 GB | 256K | 56 | Runs great |
| 8 | Qwen3 14B Alibaba · Apache 2.0 |
Good European and CJK coverage at 14B. | 14.8B | Q4_K_M · 10.2 GB | 128K | 37 | Runs great |
| 9 | Gemma 4 E4B Google · Apache 2.0 |
140+ languages on a laptop. | 8.0B | Q4_K_M · 5.2 GB | 128K | 68 | Runs great |
| 10 | Mistral NeMo 12B Mistral AI · Apache 2.0 |
Built multilingual, with a Tekken tokenizer efficient for European languages. | 12.2B | Q4_K_M · 8.7 GB | 128K | 44 | Runs great |
| 11 | Qwen3.5 4B Alibaba · Apache 2.0 |
Usable translation in 8 GB. | 4.66B | Q4_K_M · 3.5 GB | 256K | 116 | Runs great |
| 12 | Qwen3.5 2B Alibaba · Apache 2.0 |
Phrase-level translation on 4 GB. | 2.27B | Q4_K_M · 2.0 GB | 256K | 239 | Runs great |
The whole translation shortlist, and what each one needs
Every model on the list, best first, with the smallest discrete card and the smallest Mac that run it with 15% headroom at the recommended quantisation and 8K context. Every class has its own page, Macs from 8 to 512 GB included — the full list.
| # | Model | Params | Weights | Needs (GPU) | Needs (Mac) | Ollama tag |
|---|---|---|---|---|---|---|
| 1 | Qwen3.8 27B The current default local Qwen: dense 27B, text + image + video, 262K context. Only 16 of its 64 blocks keep a KV cache, so long context is cheap. |
27.8B | 15.6 GB @ Q4_K_M | 24 GB card | 32 GB Mac | qwen3.8:27b |
| 2 | Gemma 4 31B The dense flagship: strongest maths of the 24–32 GB class (89% AIME), clean prose, vision. Q4 is a tight 24 GB fit. |
31.3B | 17.6 GB @ Q4_K_M | 32 GB card | 32 GB Mac | gemma4:31b |
| 3 | Qwen3 32B The classic 24 GB target, and still the strongest local translator under 70B. Qwen3.8 27B is smaller and better at everything else. |
32.8B | 18.4 GB @ Q4_K_M | 32 GB card | 36 GB Mac | qwen3:32b |
| 4 | LLM-jp 4 33B Thinking Japan’s national-institute reasoning model, Japanese and English. A plain dense Llama-style 33B: Q4 is a tight 24 GB fit. |
33.2B | 18.7 GB @ Q4_K_M | 32 GB card | 36 GB Mac | — |
| 5 | Gemma 4 26B-A4B Mixture of experts with 3.8B active. Slower to think than Qwen3.6 35B-A3B, faster to answer, and it sees images. |
26.5B MoE | 14.9 GB @ Q4_K_M | 24 GB card | 32 GB Mac | gemma4:26b |
| 6 | Gemma 4 12B The "unified" Gemma 4: text, image and audio in one 12B that fits a 12 GB card at Q4. 140+ languages. |
12B | 6.7 GB @ Q4_K_M | 11 GB card | 16 GB Mac | gemma4:12b |
| 7 | Qwen3.5 9B The default for 8–12 GB cards in 2026: beats every older 8B on every published benchmark, with vision. |
9.65B | 5.4 GB @ Q4_K_M | 10 GB card | 16 GB Mac | qwen3.5:9b |
| 8 | Qwen3 14B The largest Qwen3 that fits a 12 GB card at Q4 with room for context. |
14.8B | 8.3 GB @ Q4_K_M | 16 GB card | 24 GB Mac | qwen3:14b |
| 9 | Gemma 4 E4B The laptop Gemma. 4.5B effective, 8B on disk; a single KV head per window layer keeps its cache tiny. |
8.0B | 4.5 GB @ Q4_K_M | 8 GB card | 16 GB Mac | gemma4:e4b |
| 10 | Mistral NeMo 12B Multilingual 12B with a 128K window, built with NVIDIA. A roleplay and fiction favourite that refuses to die. |
12.2B | 6.9 GB @ Q4_K_M | 12 GB card | 16 GB Mac | mistral-nemo:12b |
| 11 | Qwen3.5 4B The 8 GB coding agent. Q4 lands near 3.4 GB, leaving room for a real context window. |
4.66B | 2.6 GB @ Q4_K_M | 6 GB card | 8 GB Mac | qwen3.5:4b |
| 12 | Qwen3.5 2B Phone-class, and multimodal. Replaces Llama 3.2 3B as the "it runs on anything" answer. |
2.27B | 1.3 GB @ Q4_K_M | 6 GB card | 8 GB Mac | qwen3.5:2b |
As everywhere on this site, a larger model at Q4 usually beats a smaller one at Q8, and translation shows it more than most: rare words and idioms live in the parameter count. If two rows both fit, take the larger.
Other jobs