DeepSeek V4 Flash ↗
5 measured3 published
- AutomationBench · local 300GB10 · 128GB · UD-IQ3_XXS · llama.cpp19.3% ↗
- GPQA DiamondGB10 · 128GB · UD-IQ3_XXS · llama.cpp75.8% ↗
- IFBenchGB10 · 128GB · UD-IQ3_XXS · llama.cpp65.3% ↗
Find a model.
Follow the evidence.
Local measurements. Published comparisons. Find a setup for your work.
● Measured in our lab◇ Published reference
Where it runs
26 models · 79 results
GLM-5.3-Flash ↗
15 measured2 published
- AutomationBench · local 3002× GB10 (DGX Spark head + ASUS GX10 worker) · 256GB total · EXL3 TR3 4bpw · vLLM46.3% ↗
- AutomationBench · local 3002× GB10 (DGX Spark head + ASUS GX10 worker) · 256GB total · EXL3 TR3 4bpw · TensorFold48.7% ↗
- Terminal-Bench 2.12× GB10 (DGX Spark head + ASUS GX10 worker) · 256GB total · NVFP4/MXFP8 CSF-QAD · Flare69.7% ↗
Qwen 3.8 27B ↗
5 measured4 published
- AutomationBench · local 300GB10 · 128GB · UD-Q4_K_XL · llama.cpp18.3% ↗
- GPQA DiamondGB10 · 128GB · UD-Q4_K_XL · llama.cpp73.2% ↗
- IFBenchGB10 · 128GB · UD-Q4_K_XL · llama.cpp42.3% ↗
Qwen 3.8 Flash-Next ↗
5 measured0 published
- IFBenchGB10 · 128GB · UD-IQ3_XXS · llama.cpp66.0% ↗
- AutomationBench · local 300GB10 · 128GB · UD-IQ3_XXS · llama.cpp41.7% ↗
- GPQA DiamondGB10 · 128GB · UD-IQ3_XXS · llama.cpp88.4% ↗
Claude Fable 5 ↗
0 measured5 published
- GPQA DiamondCloud / API · Published92.6% ↗
- Terminal-Bench 2.1Cloud / API · Published83.1% ↗
- AutomationBench · public 600Cloud / API · Published46.2% ↗
Claude Fable 5.1 ↗
0 measured1 published
Explore model & setups →Claude Opus 4.8 ↗
0 measured6 published
- GPQA DiamondCloud / API · Published92.0% ↗
- Terminal-Bench 2.1Cloud / API · Published78.9% ↗
- GPQA DiamondCloud / API · Published93.6% ↗
Claude Opus 5 ↗
0 measured2 published
- GPQA DiamondCloud / API · Published93.7% ↗
- AutomationBench · public 600Cloud / API · Published50.3% ↗
Claude Opus 5.5 ↗
0 measured1 published
Explore model & setups →Claude Sonnet 5.5 ↗
0 measured0 publishedPublished API · no Index cite
Gemini 3.1 Pro Preview ↗
0 measured4 published
- GPQA DiamondCloud / API · Published94.3% ↗
- Terminal-Bench 2.1Cloud / API · Published70.7% ↗
- GPQA DiamondCloud / API · Published94.3% ↗
Gemini 3.8 Flash ↗
0 measured2 published
Explore model & setups →Gemini 4 Argon ↗
0 measured0 publishedPublished API · no Index cite
GLM-5.3-Flash NVFP4 ↗
0 measured3 published
- Terminal-Bench 2.1Setup unspecified · Published83.2% ↗
- IFBenchSetup unspecified · Published60.5% ↗
- GPQA DiamondSetup unspecified · Published92.1% ↗
GPT-5.6 Luna ↗
0 measured2 published
Explore model & setups →GPT-5.6 Sol ↗
0 measured5 published
- GPQA DiamondCloud / API · Published94.6% ↗
- Terminal-Bench 2.1Cloud / API · Published88.8% ↗
- Terminal-Bench 2.1Cloud / API · Published91.9% ↗
GPT-5.6 Terra ↗
0 measured3 published
- GPQA DiamondCloud / API · Published92.9% ↗
- Terminal-Bench 2.1Cloud / API · Published87.4% ↗
- AutomationBench · public 600Cloud / API · Published37.2% ↗
GPT-6 Astra ↗
0 measured2 published
Explore model & setups →GPT-6 Luna ↗
0 measured0 publishedPublished API · no Index cite
GPT-6 Sol ↗
0 measured1 published
Explore model & setups →GPT-6.1 Sol ↗
0 measured0 publishedPublished API · no Index cite
MiMo-V2.6-Distill-Qwen-9B ↗
0 measured1 publishedLocal plan
Explore model & setups →MiMo-V2.6-Flash-RL ↗
0 measured0 publishedWon't fit GB10
MiMo-V2.6-Pro-RL ↗
0 measured0 publishedWon't fit GB10
Mistral Large 4 ↗
0 measured0 publishedWon't fit GB10
Reflection Beam ↗
0 measured0 publishedWon't fit GB10