The community pulse on AI
How the models
stack up right now.
People rank the models they actually use. tier-bench combines those opinions into a clear, editable community tier list.
Overall community bench
0 contributorsS
A
B
C
D
E
F
#1Fable 5Anthropic0.00#2Mythos 5Anthropic0.00#3Opus 5Anthropic0.00#4Sonnet 5Anthropic0.00#5DeepSeekDeepSeek0.00#6Gemini ProGoogle0.00#7Gemini FlashGoogle0.00#8GLMZ.aiNew0.00#9GPT LunaOpenAI0.00#10GPT SolOpenAI0.00#11GPT TerraOpenAI0.00#12GrokxAI0.00#13InklingThinking Machines0.00#14KimiMoonshot AI0.00#15MiMoXiaomi0.00#16MistralMistral AI0.00#17MuseMeta0.00#18Qwen 27BQwen0.00#19Qwen MaxQwen0.00
Scores are community opinion, not objective benchmark results. Select any model to inspect its distribution and comments.
Chatting
Natural conversation, personality, memory of context, and enjoyable interaction.
03 / TIER LISTMath
Correct reasoning, proof quality, calculation, and explaining mathematical ideas.
04 / TIER LISTCode quality
Correctness, architecture, debugging, review quality, and maintainability.
05 / TIER LISTSteerability
Instruction following, controllable style, initiative, and course correction.
06 / TIER LISTMost value
Usefulness relative to price, limits, speed, and availability.
Request a new list
Propose a focused community tier list and let members vote on it.