LIVE
OpenAI moves toward a rigorous lattice definition of the Standard Model15/11/26 · OpenAI|Open-Weight Reasoning Model Learns to Operate Real Robots Through Simulation-Only Training13/11/26|Forensic analysis finds traces of export-controlled GPUs in Chinese frontier models12/11/26|Google DeepMind unveils Gemini 4 Hydrogen12/11/26 · Google DeepMind|AlphaGenome 2: DeepMind refines its model for decoding non-coding DNA11/11/26 · Google DeepMind|Google DeepMind rolls out Gemini 4 Carbon10/11/26 · Google DeepMind|Google DeepMind unveils AlphaProof Code for formal software verification10/11/26 · Google DeepMind|OpenAI launches a dedicated business presence feature in ChatGPT10/11/26 · OpenAI|Anthropic launches investor roadshow ahead of planned IPO10/11/26 · Anthropic|A petition calls for Jensen Huang's return as Nvidia CEO16/11/26 · Nvidia|OpenAI strengthens tools against online child sexual exploitation16/11/26 · OpenAI|OpenAI Expands Free ChatGPT Access to 200,000 Researchers16/11/26 · OpenAI|OpenAI moves toward a rigorous lattice definition of the Standard Model15/11/26 · OpenAI|Open-Weight Reasoning Model Learns to Operate Real Robots Through Simulation-Only Training13/11/26|Forensic analysis finds traces of export-controlled GPUs in Chinese frontier models12/11/26|Google DeepMind unveils Gemini 4 Hydrogen12/11/26 · Google DeepMind|AlphaGenome 2: DeepMind refines its model for decoding non-coding DNA11/11/26 · Google DeepMind|Google DeepMind rolls out Gemini 4 Carbon10/11/26 · Google DeepMind|Google DeepMind unveils AlphaProof Code for formal software verification10/11/26 · Google DeepMind|OpenAI launches a dedicated business presence feature in ChatGPT10/11/26 · OpenAI|Anthropic launches investor roadshow ahead of planned IPO10/11/26 · Anthropic|A petition calls for Jensen Huang's return as Nvidia CEO16/11/26 · Nvidia|OpenAI strengthens tools against online child sexual exploitation16/11/26 · OpenAI|OpenAI Expands Free ChatGPT Access to 200,000 Researchers16/11/26 · OpenAI|
Hardware Match

Which model for 8 × B300 · 2304 Go?

2304 GB usable for the model. 121 open-weight models fit in memory.

The margin goes to context: count ~1 GB per 8k tokens for a 7B model. A small margin means a short context.

Unlocked by this setup

These models fit on no smaller setup in the catalogue (the next one down offers 1440 usable GB), even in their lightest quantization.

ModelEloQuantizationRAM requiredMarginMemory headroomRun
Kimi K3Moonshot AI · 2800B1487Q4_K_M1758 Go+546 GotightHugging Face ↗
Kimi K3.1 ProMoonshot AI · 2800B1485MXFP41718 Go+586 GotightHugging Face ↗

Also fit on a smaller setup

ModelEloQuantizationRAM requiredMarginMemory headroomRun
GLM-5.3Zhipu AI · 753B1482FP161659 Go+645 GotightHugging Face ↗
DeepSeek V4.1-FlashDeepSeek · 552B1475FP8563 Go+1741 GocomfortableHugging Face ↗
GLM-5.2Zhipu AI · 754B1475FP161661 Go+643 GotightHugging Face ↗
GLM-5Zhipu AI · 744B1457FP161639 Go+665 GotightHugging Face ↗
Gemma 4 31BGoogle · 31B1452FP1670 Go+2234 GocomfortableHugging Face ↗
Qwen3.5 397B-A17BAlibaba · 397B1441FP16875 Go+1429 GocomfortableHugging Face ↗
DeepSeek V4-FlashDeepSeek · 284B1439Q4_K_MFP16 also fits · 627 Go180 Go+2124 Gocomfortable ollama run deepseek-v4-flash
Gemma 4 26B-A4BGoogle · 26B1437FP1659 Go+2245 GocomfortableHugging Face ↗
Qwen3.8 27BAlibaba · 27B1434FP1661 Go+2243 GocomfortableHugging Face ↗
Mistral Medium 3.5Mistral AI · 128B1426FP16284 Go+2020 GocomfortableHugging Face ↗
Nemotron 3 Ultra 550B-A55BNVIDIA · 550B1426FP161212 Go+1092 GocomfortableHugging Face ↗
Mistral Large 4Mistral AI · 1000B1420FP81168 Go+1136 GocomfortableHugging Face ↗
Kimi K2Moonshot AI · 1026B1418FP162259 Go+45 GotightHugging Face ↗
Mistral Large 3Mistral AI · 675B1415FP161487 Go+817 GocomfortableHugging Face ↗
GLM-4.5Z.ai · 355B1411FP16783 Go+1521 GocomfortableHugging Face ↗
DeepSeek-R1DeepSeek · 671B1398Q4_K_MFP16 also fits · 1478 Go423 Go+1881 Gocomfortable ollama run deepseek-r1:671b
DeepSeek V3DeepSeek · 685B1395Q4_K_MFP16 also fits · 1509 Go431 Go+1873 Gocomfortable ollama run deepseek-v3
Qwen3-Coder 480BAlibaba · 480B1387Q4_K_MFP16 also fits · 1058 Go303 Go+2001 Gocomfortable ollama run qwen3-coder:480b
Qwen3 235B-A22BAlibaba · 235B1375Q4_K_MFP16 also fits · 519 Go149 Go+2155 Gocomfortable ollama run qwen3:235b
GLM-4.5 AirZ.ai · 106B1373FP16235 Go+2069 GocomfortableHugging Face ↗
Gemma 3 27BGoogle · 27B1365Q4_K_MFP16 also fits · 61 Go19 Go+2285 Gocomfortable ollama run gemma3:27b
Nemotron 3 Super 120B-A12BNVIDIA · 120B1360FP16266 Go+2038 GocomfortableHugging Face ↗
Mistral Small 3.2Mistral AI · 24B1356Q4_K_MFP16 also fits · 55 Go17 Go+2287 Gocomfortable ollama run mistral-small3.2
gpt-oss-120bOpenAI · 117B1352MXFP473 Go+2231 Gocomfortable ollama run gpt-oss:120b
Qwen3 32BAlibaba · 32.8B1347Q4_K_MFP16 also fits · 74 Go23 Go+2281 Gocomfortable ollama run qwen3:32b
Gemma 3 12BGoogle · 12B1342Q4_K_MFP16 also fits · 28 Go9 Go+2295 Gocomfortable ollama run gemma3:12b
QwQ 32BAlibaba · 32.8B1336Q4_K_MFP16 also fits · 74 Go23 Go+2281 Gocomfortable ollama run qwq
Llama 4 MaverickMeta · 400B1327Q4_K_MFP16 also fits · 882 Go253 Go+2051 Gocomfortable ollama run llama4:maverick
Llama 4 ScoutMeta · 109B1322Q4_K_MFP16 also fits · 242 Go70 Go+2234 Gocomfortable ollama run llama4:scout
Gemma 3n E4BGoogle · 4B1318Q4_K_MFP16 also fits · 11 Go5 Go+2299 Gocomfortable ollama run gemma3n:e4b
gpt-oss-20bOpenAI · 21B1318MXFP417 Go+2287 Gocomfortable ollama run gpt-oss:20b
Llama 3.3 70BMeta · 70.6B1318Q4_K_MFP16 also fits · 157 Go46 Go+2258 Gocomfortable ollama run llama3.3:70b
Nemotron 3 Nano 30B-A3BNVIDIA · 30B1314FP1668 Go+2236 GocomfortableHugging Face ↗
Granite 4.1 8BIBM · 8B1304FP1620 Go+2284 GocomfortableHugging Face ↗
Gemma 3 4BGoogle · 4B1303Q4_K_MFP16 also fits · 11 Go5 Go+2299 Gocomfortable ollama run gemma3:4b
Qwen2.5 72BAlibaba · 72.7B1303Q4_K_MFP16 also fits · 162 Go48 Go+2256 Gocomfortable ollama run qwen2.5:72b
Llama 3.1 Nemotron 70BNVIDIA · 70.6B1299Q4_K_MFP16 also fits · 157 Go46 Go+2258 Gocomfortable ollama run nemotron:70b
Qwen2.5-Coder 32BAlibaba · 32.8B1271Q4_K_MFP16 also fits · 74 Go23 Go+2281 Gocomfortable ollama run qwen2.5-coder:32b
Aya Expanse 32BCohere · 32.3B1267Q4_K_MFP16 also fits · 73 Go22 Go+2282 Gocomfortable ollama run aya-expanse:32b
Command R+Cohere · 104B1262Q4_K_MFP16 also fits · 231 Go67 Go+2237 Gocomfortable ollama run command-r-plus
Phi-4Microsoft · 14.7B1256Q4_K_MFP16 also fits · 34 Go11 Go+2293 Gocomfortable ollama run phi4
Phi-4 ReasoningMicrosoft · 14.7B1256Q4_K_MFP16 also fits · 34 Go11 Go+2293 Gocomfortable ollama run phi4-reasoning
OLMo 2 32BAllen AI · 32B1252FP1672 Go+2232 GocomfortableHugging Face ↗
CodestralMistral AI · 22.2B1240Q4_K_MFP16 also fits · 51 Go16 Go+2288 Gocomfortable ollama run codestral
Ministral 8BMistral AI · 8.02B1238FP1620 Go+2284 GocomfortableHugging Face ↗
Llama 3.1 8BMeta · 8.03B1211Q4_K_MFP16 also fits · 20 Go7 Go+2297 Gocomfortable ollama run llama3.1:8b
Mixtral 8x7BMistral AI · 46.7B1197Q4_K_MFP16 also fits · 105 Go31 Go+2273 Gocomfortable ollama run mixtral:8x7b
Llama 3.2 3BMeta · 3.21B1167Q4_K_MFP16 also fits · 9 Go4 Go+2300 Gocomfortable ollama run llama3.2:3b
Llama 3.2 1BMeta · 1.24B1111Q4_K_MFP16 also fits · 5 Go3 Go+2301 Gocomfortable ollama run llama3.2:1b
Mistral 7BMistral AI · 7.25B1111Q4_K_MFP16 also fits · 18 Go7 Go+2297 Gocomfortable ollama run mistral:7b

Not yet ranked by the Arena

No Arena Elo linked: recent release, model absent from the Arena, or a match we have not made yet.

ModelEloQuantizationRAM requiredMarginMemory headroomRun
Qwen3 0.6BAlibaba · 0.6B—Q4_K_MFP16 also fits · 3 Go2 Go+2302 Gocomfortable ollama run qwen3:0.6b
Gemma 3 1BGoogle · 1B—Q4_K_MFP16 also fits · 4 Go3 Go+2301 Gocomfortable ollama run gemma3:1b
Phi-4 MiniMicrosoft · 3.8B—Q4_K_MFP16 also fits · 10 Go4 Go+2300 Gocomfortable ollama run phi4-mini
Qwen3 4BAlibaba · 4.02B—Q4_K_MFP16 also fits · 11 Go5 Go+2299 Gocomfortable ollama run qwen3:4b
Gemma 4 E2BGoogle · 2B—FP166 Go+2298 GocomfortableHugging Face ↗
CodeGemma 7BGoogle · 8.5B—Q4_K_MFP16 also fits · 21 Go7 Go+2297 Gocomfortable ollama run codegemma
Command R7BCohere · 8B—Q4_K_MFP16 also fits · 20 Go7 Go+2297 Gocomfortable ollama run command-r7b
DeepSeek-R1 Distill 8BDeepSeek · 8.03B—Q4_K_MFP16 also fits · 20 Go7 Go+2297 Gocomfortable ollama run deepseek-r1:8b
Granite 3.3 8BIBM · 8.1B—Q4_K_MFP16 also fits · 20 Go7 Go+2297 Gocomfortable ollama run granite3.3:8b
Hermes 3 8BNous Research · 8.03B—Q4_K_MFP16 also fits · 20 Go7 Go+2297 Gocomfortable ollama run hermes3:8b
MiniCPM5-2BOpenBMB · 2.5B—FP167 Go+2297 GocomfortableHugging Face ↗
Qwen3 8BAlibaba · 8.2B—Q4_K_MFP16 also fits · 20 Go7 Go+2297 Gocomfortable ollama run qwen3:8b
Falcon 3 10BTII · 10.3B—Q4_K_MFP16 also fits · 25 Go8 Go+2296 Gocomfortable ollama run falcon3:10b
Granite 4.1 3BIBM · 3B—FP169 Go+2295 GocomfortableHugging Face ↗
Ministral 3 3BMistral AI · 3B—FP169 Go+2295 GocomfortableHugging Face ↗
SmolLM3 3BHugging Face · 3.1B—FP169 Go+2295 GocomfortableHugging Face ↗
Mistral Nemo 12BMistral AI · 12.2B—Q4_K_MFP16 also fits · 29 Go10 Go+2294 Gocomfortable ollama run mistral-nemo
Gemma 4 E4BGoogle · 4B—FP1611 Go+2293 GocomfortableHugging Face ↗
Qwen3 14BAlibaba · 14.8B—Q4_K_MFP16 also fits · 35 Go11 Go+2293 Gocomfortable ollama run qwen3:14b
DeepSeek-Coder-V2 LiteDeepSeek · 15.7B—Q4_K_MFP16 also fits · 37 Go12 Go+2292 Gocomfortable ollama run deepseek-coder-v2:16b
StarCoder 2BigCode · 16B—Q4_K_MFP16 also fits · 37 Go12 Go+2292 Gocomfortable ollama run starcoder2:15b
Devstral SmallMistral AI · 24B—Q4_K_MFP16 also fits · 55 Go17 Go+2287 Gocomfortable ollama run devstral
Lucie 7BOpenLLM France · 6.7B—FP1617 Go+2287 GocomfortableHugging Face ↗
Magistral SmallMistral AI · 24B—Q4_K_MFP16 also fits · 55 Go17 Go+2287 Gocomfortable ollama run magistral
OLMo 3 7BAllen AI · 7B—FP1617 Go+2287 GocomfortableHugging Face ↗
Apertus 8BSwiss AI · 8B—FP1620 Go+2284 GocomfortableHugging Face ↗
Ministral 3 8BMistral AI · 8B—FP1620 Go+2284 GocomfortableHugging Face ↗
InternLM3 8BShanghai AI Lab · 8.8B—FP1621 Go+2283 GocomfortableHugging Face ↗
Apertus 1.5 8BSwiss AI Initiative · 9B—FP1622 Go+2282 GocomfortableHugging Face ↗
EuroLLM 9BUTTER Project · 9.15B—FP1622 Go+2282 GocomfortableHugging Face ↗
DeepSeek-R1 Distill 32BDeepSeek · 32.8B—Q4_K_MFP16 also fits · 74 Go23 Go+2281 Gocomfortable ollama run deepseek-r1:32b
Gemma 4 12BGoogle · 12B—FP1628 Go+2276 GocomfortableHugging Face ↗
Ministral 3 14BMistral AI · 14B—FP1633 Go+2271 GocomfortableHugging Face ↗
Qwen3.6 27BAlibaba · 27B—FP1661 Go+2243 GocomfortableHugging Face ↗
Qwen3.6 35B-A3BAlibaba · 35B—FP1679 Go+2225 GocomfortableHugging Face ↗
Apertus 1.5 70BSwiss AI Initiative · 70B—FP16156 Go+2148 GocomfortableHugging Face ↗
Kolibri 1Aleph Alpha · 78.1B—FP16174 Go+2130 GocomfortableHugging Face ↗
Mistral Small 4Mistral AI · 119B—FP16264 Go+2040 GocomfortableHugging Face ↗
Kimi K2.7 CodeMoonshot AI · 1000B—FP162202 Go+102 GotightHugging Face ↗

Image, video, audio and 3D

These models do not generate text, so the text Arena does not rank them. Indicative order, not a quality ranking.

ModelEloQuantizationRAM requiredMarginMemory headroomRun
Kokoro-82MHexgrad · 0.08B—FP162 Go+2302 GocomfortableHugging Face ↗
ChatterboxResemble AI · 0.5B—FP163 Go+2301 GocomfortableHugging Face ↗
Parakeet TDT 0.6B v3NVIDIA · 0.6B—FP163 Go+2301 GocomfortableHugging Face ↗
TripoSRStability AI · 0.5B—FP163 Go+2301 GocomfortableHugging Face ↗
Stable Fast 3DStability AI · 1B—FP164 Go+2300 GocomfortableHugging Face ↗
TRELLISMicrosoft · 1.2B—FP165 Go+2299 GocomfortableHugging Face ↗
Whisper large-v3OpenAI · 1.55B—FP165 Go+2299 GocomfortableHugging Face ↗
Dia 1.6BNari Labs · 1.6B—FP166 Go+2298 GocomfortableHugging Face ↗
Hunyuan3D 2.1Tencent · 3B—FP169 Go+2295 GocomfortableHugging Face ↗
Orpheus 3BCanopy Labs · 3B—FP169 Go+2295 GocomfortableHugging Face ↗
Voxtral Mini 3BMistral AI · 3B—FP169 Go+2295 GocomfortableHugging Face ↗
YuE2-3Bm-a-p (Multimodal Art Projection) · 3B—FP169 Go+2295 GocomfortableHugging Face ↗
FLUX.2 [klein] 4BBlack Forest Labs · 4B—FP1611 Go+2293 GocomfortableHugging Face ↗
TRELLIS.2Microsoft · 4B—FP1611 Go+2293 GocomfortableHugging Face ↗
CogVideoX 5BZhipu AI · 5B—FP1613 Go+2291 GocomfortableHugging Face ↗
Wan 2.2 5BAlibaba · 5B—FP1613 Go+2291 GocomfortableHugging Face ↗
Ming-Image 0.1 DesignAnt Group · 6B—FP1615 Go+2289 GocomfortableHugging Face ↗
NitroFusion2SketchX (University of Surrey) · 6B—FP1615 Go+2289 GocomfortableHugging Face ↗
Z-Image TurboAlibaba · 6B—FP1615 Go+2289 GocomfortableHugging Face ↗
HunyuanVideo 1.5Tencent · 8.3B—FP1620 Go+2284 GocomfortableHugging Face ↗
GLM-ImageZhipu AI · 9B—FP1622 Go+2282 GocomfortableHugging Face ↗
Mochi 1Genmo · 10B—FP1624 Go+2280 GocomfortableHugging Face ↗
FLUX.1 [dev]Black Forest Labs · 12B—FP1628 Go+2276 GocomfortableHugging Face ↗
HiDream-I1HiDream · 17B—FP1639 Go+2265 GocomfortableHugging Face ↗
Qwen-Image 2512Alibaba · 20B—FP1646 Go+2258 GocomfortableHugging Face ↗
LTX-2.3Lightricks · 22B—FP1650 Go+2254 GocomfortableHugging Face ↗
Voxtral Small 24BMistral AI · 24B—FP1655 Go+2249 GocomfortableHugging Face ↗
Wan 2.2 A14BAlibaba · 27B—FP1661 Go+2243 GocomfortableHugging Face ↗
FLUX.2 [dev]Black Forest Labs · 32B—FP1672 Go+2232 GocomfortableHugging Face ↗
HunyuanImage 3.0Tencent · 80B—FP16178 Go+2126 GocomfortableHugging Face ↗

An “ollama run” command without a suffix pulls the Q4_K_M build from the Ollama library: the row then shows that quantization and its memory, and flags whether FP16 or Q8_0 would also fit. Exception: a model published in a native format (MXFP4 for gpt-oss, FP8 for DeepSeek V4.1) exists in one build only, which is what the command loads. via Hugging Face or an explicit Ollama tag.

Share this verdict

One link, the same shortlist, the same reasons.