AutoBench Run 5 - December 2025

Latest AutoBench run with models Gpt 5.2, Claude Opus 4.5, DeepSeek 3.2 Speciale and more

Latest

Date

December 16, 2025

Version

2025-12-16

Models

New Models

Run data

Overall Comparison Costs Latency P99 Latency Domain

AutoBench provides a complete dashboard with highly customizable domain-specific average ranks and precise efficiency metrics (average answer cost, average answer duration, P99 Latency). For further detail also check our Hugging Face Leaderboard.

Model	Score	Avg Cost ($ Cents)	Avg Latency (sec)	P99 Latency (sec)	Iterations
Gpt-5.2-pro	4.48 (#1)	81.88 (#35)	261s (#33)	784s (#33)	303
Gpt-5.2	4.43 (#2)	7.36 (#30)	130s (#23)	434s (#25)	312
gemini-3-pro-preview	4.41 (#3)	6.85 (#29)	76s (#12)	186s (#9)	312
Claude-opus-4.5	4.39 (#4)	17.26 (#34)	144s (#25)	373s (#22)	313
gpt-5.1	4.38 (#5)	10.80 (#32)	227s (#31)	627s (#30)	310
Kimi-K2-thinking	4.32 (#6)	1.86 (#24)	248s (#32)	729s (#32)	287
claude-sonnet-4.5	4.3 (#7)	11.39 (#33)	170s (#27)	477s (#27)	307
gemini-2.5-pro	4.29 (#9)	6.48 (#28)	87s (#16)	222s (#13)	313
gpt-5-mini	4.29 (#8)	0.91 (#20)	93s (#18)	258s (#16)	312
grok-4.1-fast-thinking	4.21 (#10)	0.27 (#10)	69s (#10)	207s (#11)	306
Qwen3-235B-A22B-Thinking-2507	4.2 (#11)	0.32 (#11)	317s (#35)	811s (#34)	283
grok-4	4.2 (#12)	8.12 (#31)	180s (#29)	562s (#29)	293
gpt-oss-120b	4.18 (#13)	0.11 (#6)	75s (#11)	292s (#18)	292
claude-haiku-4.5	4.17 (#15)	3.79 (#26)	111s (#21)	317s (#19)	312
gemini-2.5-flash	4.17 (#14)	2.12 (#25)	66s (#8)	174s (#7)	312
Deepseek-v3.2-speciale	4.14 (#16)	0.47 (#15)	310s (#34)	833s (#35)	288
GLM-4.6	4.13 (#17)	1.25 (#22)	187s (#30)	630s (#31)	306
deepSeek-R1-0528	4.12 (#18)	0.99 (#21)	171s (#28)	477s (#28)	308
Deepseek-v3.2	4.11 (#19)	0.09 (#5)	125s (#22)	410s (#23)	311
Kimi-K2-0905	4.11 (#20)	0.33 (#12)	83s (#15)	329s (#20)	312
Nova-2-lite-v1	4.06 (#22)	3.94 (#27)	61s (#7)	132s (#3)	277
gpt-5-nano	4.06 (#21)	0.34 (#13)	100s (#19)	269s (#17)	309
Qwen3-next-80b-a3b-thinking	4.03 (#23)	0.75 (#19)	78s (#14)	227s (#14)	312
Minimax-m2	3.99 (#24)	0.71 (#18)	137s (#24)	473s (#26)	308
Qwen3-235B-A22B-2507	3.98 (#25)	0.19 (#8)	105s (#20)	337s (#21)	302
gemini-2.5-flash-lite	3.95 (#26)	0.21 (#9)	20s (#1)	69s (#2)	313
Mistral-large-2512	3.94 (#27)	0.51 (#16)	90s (#17)	198s (#10)	307
grok-4.1-fast	3.88 (#28)	0.08 (#3)	24s (#2)	57s (#1)	312
GLM-4.5-Air	3.86 (#29)	0.54 (#17)	163s (#26)	425s (#24)	306
Mistral-medium-3.1	3.81 (#30)	0.38 (#14)	52s (#6)	147s (#5)	306
Gpt-oss-20b	3.78 (#31)	0.07 (#2)	39s (#4)	183s (#8)	310
llama-3.3-nemotron-super-49b-v1.5	3.78 (#32)	0.18 (#7)	76s (#13)	240s (#15)	311
Ministral-8b-2512	3.57 (#33)	0.05 (#1)	31s (#3)	154s (#6)	306
nemotron-nano-9b-v2	3.5 (#34)	0.08 (#4)	67s (#9)	212s (#12)	311
nova-premier-v1	3.47 (#35)	1.30 (#23)	52s (#5)	135s (#4)	312