Only the same benchmark, version, workload, scenario and unit share a row. Missing results are not zero. System results include the full submitted hardware and software configuration.
Arena text preference / chinesehuman preference · 2026-10-28 · rating
phi-4
1,211.40 rating
Confidence interval: 1,195.86 – 1,226.93
1503 votes
Category
chinese
Evaluated configuration
Phi-4
Evaluation type
human preference
Arena text preference / codinghuman preference · 2026-10-28 · rating
phi-4
1,231.03 rating
Confidence interval: 1,221.19 – 1,240.87
3305 votes
Category
coding
Evaluated configuration
Phi-4
Evaluation type
human preference
Arena text preference / creative_writinghuman preference · 2026-10-28 · rating
phi-4
1,181.29 rating
Confidence interval: 1,171.64 – 1,190.94
4062 votes
Category
creative_writing
Evaluated configuration
Phi-4
Evaluation type
human preference
Arena text preference / englishhuman preference · 2026-10-28 · rating
phi-4
1,230.96 rating
Confidence interval: 1,225.20 – 1,236.72
14101 votes
Category
english
Evaluated configuration
Phi-4
Evaluation type
human preference
Arena text preference / exclude_tieshuman preference · 2026-10-28 · rating
phi-4
1,100.08 rating
Confidence interval: 1,092.39 – 1,107.76
15789 votes
Category
exclude_ties
Evaluated configuration
Phi-4
Evaluation type
human preference
Arena text preference / experthuman preference · 2026-10-28 · rating
phi-4
1,202.60 rating
Confidence interval: 1,185.38 – 1,219.82
1124 votes
Category
expert
Evaluated configuration
Phi-4
Evaluation type
human preference
Arena text preference / frenchhuman preference · 2026-10-28 · rating
phi-4
1,224.51 rating
Confidence interval: 1,184.88 – 1,264.15
205 votes
Category
french
Evaluated configuration
Phi-4
Evaluation type
human preference
Arena text preference / germanhuman preference · 2026-10-28 · rating
phi-4
1,222.38 rating
Confidence interval: 1,196.99 – 1,247.78
523 votes
Category
german
Evaluated configuration
Phi-4
Evaluation type
human preference
Arena text preference / hard_promptshuman preference · 2026-10-28 · rating
phi-4
1,219.45 rating
Confidence interval: 1,211.69 – 1,227.21
5747 votes
Category
hard_prompts
Evaluated configuration
Phi-4
Evaluation type
human preference
Arena text preference / hard_prompts_englishhuman preference · 2026-10-28 · rating
phi-4
1,229.85 rating
Confidence interval: 1,220.46 – 1,239.23
3804 votes
Category
hard_prompts_english
Evaluated configuration
Phi-4
Evaluation type
human preference
Arena text preference / industry_business_and_management_and_financial_operationshuman preference · 2026-10-28 · rating
No catalogued Local AI system meets this memory estimate with the reserved headroom.
Shared-memory systems use a planning budget of 75% of installed memory, leaving 25% for the OS and other allocations. This is an assumption, not a measured GPU allocation. Firmware and driver limits may be lower and may require configuration. Memory matches do not verify a runtime, quantized model artifact or inference speed.
Includes FP16 KV cache, prompt plus generated tokens, quantization metadata and a runtime allowance. GPU matches reserve 10% of VRAM. Memory fit is not a speed or software-compatibility guarantee; quantized weights and runtime support must be available.