Updated Oct 8, 2026

SI models, measured.

Independent benchmarks, real cost per task and specs for 29 current models from 12 labs — so you can pick the right one, not just the newest.

models: 29models
labs: 12labs
benchmarks: 17benchmarks

Best for…

The current leader for each common need, straight from the data. Open any card for the full ranking.

Most capable overall
Claude Opus 5.5
Anthropic
57.6Artificial Analysis Intelligence Index
1. 2Claude Sonnet 5.556
2. 3Claude Fable 5.153.4
The best single independent estimate of general capability.
Full ranking
Coding agents
Claude Opus 5.5
Anthropic
64.8%Terminal-Bench 4.0
1. 2Claude Sonnet 5.561.8%
2. 3GPT-6 Astra58.2%
Finishes the most real terminal jobs end to end.
Full ranking
Professional work
Claude Opus 5.5
Anthropic
1,866GDPval-AA
1. 2Claude Sonnet 5.51,839
2. 3Claude Fable 5.11,756
Wins the most head-to-head comparisons on real work deliverables.
Full ranking
Novel reasoning
GPT-6 Astra
OpenAI
95%ARC-AGI-2
1. 2GPT-6.1 Sol94.2%
2. 3Claude Opus 5.593.3%
Best at working out rules it has never seen before.
Full ranking
People's choice
Gemini 4 Argon
Google
1,525Arena Text
1. 2Claude Opus 5.51,504
2. 3Claude Fable 5.11,501
The answer people preferred most often in blind votes.
Full ranking
Value near the frontier
GPT-6.1 Sol
OpenAI
71.6Capability per dollar
1. 2Gemini 4 Argon26.4
2. 3GPT-6 Astra16.2
Most capability per dollar among models within 15% of the top overall score.
Full ranking
Fastest capable model
Claude Haiku 5.5
Anthropic
242 t/sOutput speed
1. 2Muse Spark 1.3209 t/s
2. 3Claude Sonnet 5.5129 t/s
Quickest output among models scoring at least 70% of the top overall score.
Full ranking
Best open weights
MiMo-V2.6-Pro
Xiaomi
46.3Artificial Analysis Intelligence Index
1. 2GLM-5.344.8
2. 3Kimi K343.6
The strongest model you can download and run yourself.
Full ranking

Released in the last 60 days

Leaderboard

29 models · ranked by Artificial Analysis Intelligence Index

Ranked by the Artificial Analysis Intelligence Index, an independent composite of ten evaluations. A sound first cut when you do not yet know what you need.

Rank 1
Claude Opus 5.557.6
Anthropic $4 / $20New
Rank 2
Claude Sonnet 5.556
Anthropic $2 / $10New
Rank 3
Claude Fable 5.153.4
Anthropic $10 / $50New
Rank 4
GPT-6 Astra52.7
OpenAI $10 / $50New
Rank 5
Gemini 4 Argon52.6
Google $2 / $10NewPreview
Rank 6
GPT-6.1 Sol51.8
OpenAI $2 / $10New
Rank 7
Muse Spark 1.348.1
Meta $1.25 / $4.25New
Rank 8
Grok 4.746.5
SpaceXAI $2 / $6New
Rank 9
MiMo-V2.6-Pro46.3
Xiaomi $0.435 / $0.87NewOpen weights
Rank 10
Qwen3.8-Max45.4
Alibaba $2 / $6
Rank 11
GLM-5.344.8
Z.ai $1.40 / $4.40NewOpen weights
Rank 12
Kimi K343.6
Moonshot AI $3 / $15Open weights
Rank 13
Claude Haiku 5.543.4
Anthropic $0.10 / $0.50New
Rank 14
GLM-5.3-Flash41.8
Z.ai $0.15 / $0.50NewOpen weights
Rank 15
Gemini 3.8 Flash40.9
Google $0.75 / $3.75New
Rank 16
Qwen3.8-Flash-Next39.8
Alibaba $0.15 / $0.47NewOpen weights
Rank 17
DeepSeek-V4.1-Flash39.5
DeepSeek $0.30 / $1.20NewOpen weights
Rank 18
Mistral Large 438.4
Mistral AI $1.36 / $4.18NewPreview
Rank 19
GPT-6 Luna38.1
OpenAI $0.10 / $0.50New
Rank 20
DeepSeek-V4-Pro36
DeepSeek $1.32 / $3.96NewOpen weights
Rank 21
Qwen3.8-27B33.7
Alibaba $0.50 / $3NewOpen weights
Rank 22
Gemini 3.1 Pro29.7
Google $2 / $12Preview
Rank 23
MiniMax M329.2
MiniMax $0.30 / $1.20Open weights
Rank 24
Kimi K2.7 Code25.8
Moonshot AI $0.95 / $4Open weights
Rank 25
Gemini 3.5 Flash-Lite22.2
Google $0.30 / $2.50
Rank 26
Muse Glimmer 30B17.5
Meta Self-hostedNewOpen weights
Rank 27
Gemma 4 31B14.7
Google Self-hostedOpen weights
Rank 28
Mistral Medium 3.514.2
Mistral AI $1.50 / $7.50Open weights
Rank 29
Mistral Small 411.3
Mistral AI $0.15 / $0.60Open weights

Models ranked by Artificial Analysis Intelligence Index, best first. Column headers re-sort the table.
#	Model
1	Claude Opus 5.5New Anthropic	57.6	167.3	83.2%	1,504	$5.98	95 t/s	1M
2	Claude Sonnet 5.5New Anthropic	56	165	77.8%	1,471	$5.46	129 t/s	1M
3	Claude Fable 5.1New Anthropic	53.4	164.7	83.4%	1,501	$7.63	66 t/s	1M
4	GPT-6 AstraNew OpenAI	52.7	166.4	82.2%	1,477	$3.26	48 t/s	1.1M
5	Gemini 4 ArgonNewPreview Google	52.6	Not reported	Not reported	1,525	$1.99	Not reported	1M
6	GPT-6.1 SolNew OpenAI	51.8	166.1	81.6%	1,483	$0.72	55 t/s	1.1M
7	Muse Spark 1.3New Meta	48.1	156.8	81.6%	1,494	$1.60	209 t/s	1M
8	Grok 4.7New SpaceXAI	46.5	153.5	77.4%	1,442	$3.74	73 t/s	500K
9	MiMo-V2.6-ProNewOpen weights Xiaomi	46.3	Not reported	Not reported	1,480	$0.13	38 t/s	1M
10	Qwen3.8-Max Alibaba	45.4	156.4	78.5%	1,482	$5.41	37 t/s	1M
11	GLM-5.3NewOpen weights Z.ai	44.8	155.6	76.1%	1,478	$2.01	79 t/s	1M
12	Kimi K3Open weights Moonshot AI	43.6	157.4	79.2%	1,488	$2	42 t/s	1M
13	Claude Haiku 5.5New Anthropic	43.4	Not reported	72.1%	Not reported	$0.21	242 t/s	1M
14	GLM-5.3-FlashNewOpen weights Z.ai	41.8	151.9	71.6%	1,473	$0.25	51 t/s	1M
15	Gemini 3.8 FlashNew Google	40.9	156.7	75.8%	1,495	$1.24	125 t/s	1M
16	Qwen3.8-Flash-NextNewOpen weights Alibaba	39.8	Not reported	76.2%	Not reported	$0.37	56 t/s	262K
17	DeepSeek-V4.1-FlashNewOpen weights DeepSeek	39.5	154.9	81.1%	1,474	$0.27	216 t/s	1M
18	Mistral Large 4NewPreview Mistral AI	38.4	Not reported	71.8%	Not reported	$1.13	116 t/s	1M
19	GPT-6 LunaNew OpenAI	38.1	156.3	72%	1,443	$0.07	129 t/s	1.1M
20	DeepSeek-V4-ProNewOpen weights DeepSeek	36	155.3	77.4%	1,464	$0.67	100 t/s	1M
21	Qwen3.8-27BNewOpen weights Alibaba	33.7	Not reported	75.3%	1,438	$1.01	47 t/s	262K
22	Gemini 3.1 ProPreview Google	29.7	Not reported	77%	1,487	$1.30	122 t/s	1M
23	MiniMax M3Open weights MiniMax	29.2	146.9	67.3%	1,440	$0.51	94 t/s	1M
24	Kimi K2.7 CodeOpen weights Moonshot AI	25.8	Not reported	68.4%	Not reported	$0.54	84 t/s	262K
25	Gemini 3.5 Flash-Lite Google	22.2	Not reported	63.9%	1,455	$0.19	346 t/s	1M
26	Muse Glimmer 30BNewOpen weights Meta	17.5	Not reported	Not reported	1,425	$0.06	89 t/s	131K
27	Gemma 4 31BOpen weights Google	14.7	Not reported	Not reported	1,453	$0	35 t/s	256K
28	Mistral Medium 3.5Open weights Mistral AI	14.2	Not reported	Not reported	1,427	$0.50	168 t/s	256K
29	Mistral Small 4Open weights Mistral AI	11.3	Not reported	Not reported	Not reported	$0.02	160 t/s	256K

Yardstick

Updated Oct 8, 2026

SI models, measured.

Independent benchmarks, real cost per task and specs for 29 current models from 12 labs — so you can pick the right one, not just the newest.

models: 29models
labs: 12labs
benchmarks: 17benchmarks

Best for…

The current leader for each common need, straight from the data. Open any card for the full ranking.

Most capable overall
Claude Opus 5.5
Anthropic
57.6Artificial Analysis Intelligence Index
1. 2Claude Sonnet 5.556
2. 3Claude Fable 5.153.4
The best single independent estimate of general capability.
Full ranking
Coding agents
Claude Opus 5.5
Anthropic
64.8%Terminal-Bench 4.0
1. 2Claude Sonnet 5.561.8%
2. 3GPT-6 Astra58.2%
Finishes the most real terminal jobs end to end.
Full ranking
Professional work
Claude Opus 5.5
Anthropic
1,866GDPval-AA
1. 2Claude Sonnet 5.51,839
2. 3Claude Fable 5.11,756
Wins the most head-to-head comparisons on real work deliverables.
Full ranking
Novel reasoning
GPT-6 Astra
OpenAI
95%ARC-AGI-2
1. 2GPT-6.1 Sol94.2%
2. 3Claude Opus 5.593.3%
Best at working out rules it has never seen before.
Full ranking
People's choice
Gemini 4 Argon
Google
1,525Arena Text
1. 2Claude Opus 5.51,504
2. 3Claude Fable 5.11,501
The answer people preferred most often in blind votes.
Full ranking
Value near the frontier
GPT-6.1 Sol
OpenAI
71.6Capability per dollar
1. 2Gemini 4 Argon26.4
2. 3GPT-6 Astra16.2
Most capability per dollar among models within 15% of the top overall score.
Full ranking
Fastest capable model
Claude Haiku 5.5
Anthropic
242 t/sOutput speed
1. 2Muse Spark 1.3209 t/s
2. 3Claude Sonnet 5.5129 t/s
Quickest output among models scoring at least 70% of the top overall score.
Full ranking
Best open weights
MiMo-V2.6-Pro
Xiaomi
46.3Artificial Analysis Intelligence Index
1. 2GLM-5.344.8
2. 3Kimi K343.6
The strongest model you can download and run yourself.
Full ranking

Released in the last 60 days

Leaderboard

29 models · ranked by Artificial Analysis Intelligence Index

Ranked by the Artificial Analysis Intelligence Index, an independent composite of ten evaluations. A sound first cut when you do not yet know what you need.

Rank 1
Claude Opus 5.557.6
Anthropic $4 / $20New
Rank 2
Claude Sonnet 5.556
Anthropic $2 / $10New
Rank 3
Claude Fable 5.153.4
Anthropic $10 / $50New
Rank 4
GPT-6 Astra52.7
OpenAI $10 / $50New
Rank 5
Gemini 4 Argon52.6
Google $2 / $10NewPreview
Rank 6
GPT-6.1 Sol51.8
OpenAI $2 / $10New
Rank 7
Muse Spark 1.348.1
Meta $1.25 / $4.25New
Rank 8
Grok 4.746.5
SpaceXAI $2 / $6New
Rank 9
MiMo-V2.6-Pro46.3
Xiaomi $0.435 / $0.87NewOpen weights
Rank 10
Qwen3.8-Max45.4
Alibaba $2 / $6
Rank 11
GLM-5.344.8
Z.ai $1.40 / $4.40NewOpen weights
Rank 12
Kimi K343.6
Moonshot AI $3 / $15Open weights
Rank 13
Claude Haiku 5.543.4
Anthropic $0.10 / $0.50New
Rank 14
GLM-5.3-Flash41.8
Z.ai $0.15 / $0.50NewOpen weights
Rank 15
Gemini 3.8 Flash40.9
Google $0.75 / $3.75New
Rank 16
Qwen3.8-Flash-Next39.8
Alibaba $0.15 / $0.47NewOpen weights
Rank 17
DeepSeek-V4.1-Flash39.5
DeepSeek $0.30 / $1.20NewOpen weights
Rank 18
Mistral Large 438.4
Mistral AI $1.36 / $4.18NewPreview
Rank 19
GPT-6 Luna38.1
OpenAI $0.10 / $0.50New
Rank 20
DeepSeek-V4-Pro36
DeepSeek $1.32 / $3.96NewOpen weights
Rank 21
Qwen3.8-27B33.7
Alibaba $0.50 / $3NewOpen weights
Rank 22
Gemini 3.1 Pro29.7
Google $2 / $12Preview
Rank 23
MiniMax M329.2
MiniMax $0.30 / $1.20Open weights
Rank 24
Kimi K2.7 Code25.8
Moonshot AI $0.95 / $4Open weights
Rank 25
Gemini 3.5 Flash-Lite22.2
Google $0.30 / $2.50
Rank 26
Muse Glimmer 30B17.5
Meta Self-hostedNewOpen weights
Rank 27
Gemma 4 31B14.7
Google Self-hostedOpen weights
Rank 28
Mistral Medium 3.514.2
Mistral AI $1.50 / $7.50Open weights
Rank 29
Mistral Small 411.3
Mistral AI $0.15 / $0.60Open weights

Models ranked by Artificial Analysis Intelligence Index, best first. Column headers re-sort the table.
#	Model
1	Claude Opus 5.5New Anthropic	57.6	167.3	83.2%	1,504	$5.98	95 t/s	1M
2	Claude Sonnet 5.5New Anthropic	56	165	77.8%	1,471	$5.46	129 t/s	1M
3	Claude Fable 5.1New Anthropic	53.4	164.7	83.4%	1,501	$7.63	66 t/s	1M
4	GPT-6 AstraNew OpenAI	52.7	166.4	82.2%	1,477	$3.26	48 t/s	1.1M
5	Gemini 4 ArgonNewPreview Google	52.6	Not reported	Not reported	1,525	$1.99	Not reported	1M
6	GPT-6.1 SolNew OpenAI	51.8	166.1	81.6%	1,483	$0.72	55 t/s	1.1M
7	Muse Spark 1.3New Meta	48.1	156.8	81.6%	1,494	$1.60	209 t/s	1M
8	Grok 4.7New SpaceXAI	46.5	153.5	77.4%	1,442	$3.74	73 t/s	500K
9	MiMo-V2.6-ProNewOpen weights Xiaomi	46.3	Not reported	Not reported	1,480	$0.13	38 t/s	1M
10	Qwen3.8-Max Alibaba	45.4	156.4	78.5%	1,482	$5.41	37 t/s	1M
11	GLM-5.3NewOpen weights Z.ai	44.8	155.6	76.1%	1,478	$2.01	79 t/s	1M
12	Kimi K3Open weights Moonshot AI	43.6	157.4	79.2%	1,488	$2	42 t/s	1M
13	Claude Haiku 5.5New Anthropic	43.4	Not reported	72.1%	Not reported	$0.21	242 t/s	1M
14	GLM-5.3-FlashNewOpen weights Z.ai	41.8	151.9	71.6%	1,473	$0.25	51 t/s	1M
15	Gemini 3.8 FlashNew Google	40.9	156.7	75.8%	1,495	$1.24	125 t/s	1M
16	Qwen3.8-Flash-NextNewOpen weights Alibaba	39.8	Not reported	76.2%	Not reported	$0.37	56 t/s	262K
17	DeepSeek-V4.1-FlashNewOpen weights DeepSeek	39.5	154.9	81.1%	1,474	$0.27	216 t/s	1M
18	Mistral Large 4NewPreview Mistral AI	38.4	Not reported	71.8%	Not reported	$1.13	116 t/s	1M
19	GPT-6 LunaNew OpenAI	38.1	156.3	72%	1,443	$0.07	129 t/s	1.1M
20	DeepSeek-V4-ProNewOpen weights DeepSeek	36	155.3	77.4%	1,464	$0.67	100 t/s	1M
21	Qwen3.8-27BNewOpen weights Alibaba	33.7	Not reported	75.3%	1,438	$1.01	47 t/s	262K
22	Gemini 3.1 ProPreview Google	29.7	Not reported	77%	1,487	$1.30	122 t/s	1M
23	MiniMax M3Open weights MiniMax	29.2	146.9	67.3%	1,440	$0.51	94 t/s	1M
24	Kimi K2.7 CodeOpen weights Moonshot AI	25.8	Not reported	68.4%	Not reported	$0.54	84 t/s	262K
25	Gemini 3.5 Flash-Lite Google	22.2	Not reported	63.9%	1,455	$0.19	346 t/s	1M
26	Muse Glimmer 30BNewOpen weights Meta	17.5	Not reported	Not reported	1,425	$0.06	89 t/s	131K
27	Gemma 4 31BOpen weights Google	14.7	Not reported	Not reported	1,453	$0	35 t/s	256K
28	Mistral Medium 3.5Open weights Mistral AI	14.2	Not reported	Not reported	1,427	$0.50	168 t/s	256K
29	Mistral Small 4Open weights Mistral AI	11.3	Not reported	Not reported	Not reported	$0.02	160 t/s	256K

SI models, measured.

Best for…

Claude Opus 5.5

Claude Opus 5.5

Claude Opus 5.5

GPT-6 Astra

Gemini 4 Argon

GPT-6.1 Sol

Claude Haiku 5.5

MiMo-V2.6-Pro

Released in the last 60 days

Leaderboard

Filters

SI models, measured.

Best for…

Claude Opus 5.5

Claude Opus 5.5

Claude Opus 5.5

GPT-6 Astra

Gemini 4 Argon

GPT-6.1 Sol

Claude Haiku 5.5

MiMo-V2.6-Pro

Released in the last 60 days

Leaderboard

Filters