Compare models

Higher is better on every test, and the best published score in each column is highlighted. A "—" means no score has been published yet. It never means the model scored zero.

What do these tests measure?

SWE-bench Verified: Hands the AI bugs from actual software projects and counts how many it fixes. Like a coding job interview, but with real work.
SWE-bench Pro: The harder version of the coding test. Bigger codebases, trickier bugs. Scores drop for everyone, so the gaps between models become clearer.
GPQA Diamond: PhD-level science questions written so you cannot just Google the answer. Tests whether the model can reason about hard science.
Terminal-Bench 2.1: Puts the AI in front of a computer terminal and asks it to finish multi-step tasks on its own. Measures how good an "AI agent" it is.
Humanity's Last Exam: PhD-level questions across many subjects. Tests deep reasoning on the hardest questions humans can ask.
GDPval-AA v2: Real paid work from 44 different jobs, such as law, nursing, and software. Judges compare two answers side by side without knowing which model wrote them, and the winner gains rating points. A typical human expert scores 1000, so a higher number means the work was picked over a human more often.
Artificial Analysis Intelligence Index: A frequently refreshed overall score made from nine modern tests: real work, tool use, terminal tasks, science, hard questions, and long-context reasoning. It is a scorecard rather than a percent correct.

Showing 23 of 23 models

Claude Opus 5

Released 7/24/2026

Anthropicflagship

Context:1M

Price:$5

🧠 reasoning🌐 web👁️ vision

Top benchmarks:

gdpval aa1861.0 pts

swe bench verified97.0%

gpqa diamond93.4%

Claude Fable 5

Released 6/9/2026

Anthropicflagship

Context:1M

Price:$10

🧠 reasoning🌐 web👁️ vision

Top benchmarks:

gdpval aa1759.6 pts

swe bench verified95.0%

gpqa diamond92.6%

Claude Opus 4.8

Released 5/28/2026

Anthropicflagship

Context:1M

Price:$5

🧠 reasoning🌐 web👁️ vision

Top benchmarks:

gdpval aa1600.1 pts

gpqa diamond93.6%

swe bench verified88.6%

Claude Sonnet 5

Released 6/30/2026

Anthropicbalanced

Context:1M

Price:$2

🧠 reasoning🌐 web👁️ vision

Top benchmarks:

gdpval aa1603.0 pts

gpqa diamond91.1%

swe bench verified82.1%

Claude Haiku 4.5

Released 10/15/2025

Anthropicfast

Context:200K

Price:$1

🧠 reasoning🌐 web👁️ vision

Top benchmarks:

gdpval aa911.0 pts

swe bench verified73.3%

GPT-5.6 Sol

Released 7/9/2026

OpenAIflagship

Context:1.05M

Price:$5

🧠 reasoning🌐 web👁️ vision

Top benchmarks:

gdpval aa1747.8 pts

swe bench verified96.2%

gpqa diamond94.6%

GPT-5.6 Terra

Released 7/9/2026

OpenAIbalanced

Context:1.05M

Price:$2.50

🧠 reasoning🌐 web👁️ vision

Top benchmarks:

gdpval aa1593.0 pts

gpqa diamond92.9%

terminal bench87.4%

GPT-5.6 Luna

Released 7/9/2026

OpenAIfast

Context:1.05M

Price:$1

🧠 reasoning🌐 web👁️ vision

Top benchmarks:

gdpval aa1591.8 pts

swe bench verified93.0%

gpqa diamond92.3%

Gemini 3.1 Pro

Released 2/19/2026

Googleflagship

Context:1M

Price:$2

🧠 reasoning🌐 web👁️ vision

Top benchmarks:

gdpval aa962.3 pts

gpqa diamond94.3%

swe bench verified80.6%

Gemini 3.6 Flash

Released 7/21/2026

Googlebalanced

Context:1.05M

Price:$1.50

🧠 reasoning🌐 web👁️ vision

Top benchmarks:

gdpval aa1423.0 pts

terminal bench78.0%

swe bench pro58.7%

Gemini 3.5 Flash-Lite

Released 7/21/2026

Googlefast

Context:1.05M

Price:$0.30

🧠 reasoning🌐 web👁️ vision

Top benchmarks:

gdpval aa1139.0 pts

swe bench pro54.2%

terminal bench54.0%

Gemini 3.5 Flash

Released 5/19/2026

Googlefast

Context:1M

Price:$1.50

🧠 reasoning🌐 web👁️ vision

Top benchmarks:

gdpval aa1348.8 pts

gpqa diamond92.2%

swe bench verified78.0%

Grok 4.5

Released 7/16/2026

xAIflagship

Context:500K

Price:$2

🧠 reasoning🌐 web👁️ vision

Top benchmarks:

gdpval aa1527.0 pts

gpqa diamond93.1%

swe bench verified86.6%

Grok 4.1 Fast

Released 11/19/2025

xAIfast

Context:2M

Price:$0.20

🧠 reasoning🌐 web

Top benchmarks:

gpqa diamond85.3%

swe bench pro70.0%

hle17.6%

Muse Spark 1.1

Released 7/9/2026

Metaflagship

Context:1M

Price:$1.25

🧠 reasoning🌐 web👁️ vision

Top benchmarks:

gdpval aa1375.0 pts

gpqa diamond88.4%

swe bench verified82.0%

Kimi K3

Released 7/16/2026

Moonshot AIflagship

Context:1M

Price:$3

🧠 reasoning🌐 web👁️ vision

Top benchmarks:

gdpval aa1686.0 pts

gpqa diamond93.5%

swe bench verified93.4%

Inkling

Released 7/15/2026

Thinking Machinesflagship

Context:1M

Price:—

🧠 reasoning👁️ visionopen

Top benchmarks:

gdpval aa1237.0 pts

gpqa diamond87.2%

swe bench verified77.6%

Mistral Medium 3.5

Released 4/28/2026

Mistral AIbalanced

Context:256K

Price:—

🧠 reasoning🌐 web👁️ visionopen

Top benchmarks:

aa intelligence index30.0 pts

terminal bench15.6%

GLM-5.2

Released 6/13/2026

Z.ai (GLM)flagship

Context:1M

Price:—

🧠 reasoningopen

Top benchmarks:

gdpval aa1510.0 pts

gpqa diamond91.2%

terminal bench81.0%

DeepSeek V4 Pro

Released 4/24/2026

DeepSeekflagship

Context:—

Price:—

🧠 reasoningopen

Top benchmarks:

gdpval aa1306.0 pts

gpqa diamond90.1%

swe bench verified80.6%

Qwen 3.6

Released 4/16/2026

Alibaba (Qwen)flagship

Context:—

Price:—

🧠 reasoning👁️ visionopen

Top benchmarks:

gdpval aa1139.0 pts

gpqa diamond86.0%

swe bench verified73.4%

Llama 4 Maverick

Released 4/5/2025

Metaflagship

Context:1M

Price:—

👁️ visionopen

Top benchmarks:

gpqa diamond69.8%

gdpval aa7.0 pts

Llama 4 Scout

Released 4/5/2025

Metabalanced

Context:10M

Price:—

👁️ visionopen

Top benchmarks:

gdpval aa111.0 pts

gpqa diamond57.2%


Claude Opus 5 Jul 2026 Anthropicreasoningweb		no published score						$5 / $25	1M
Claude Fable 5 Jun 2026 Anthropicreasoningweb								$10 / $50	1M
Claude Opus 4.8 May 2026 Anthropicreasoningweb							no published score	$5 / $25	1M
Claude Sonnet 5 Jun 2026 Anthropicreasoningweb								$2 / $10	1M
Claude Haiku 4.5 Oct 2025 Anthropicreasoningweb		no published score	no published score	no published score	no published score		no published score	$1 / $5	200K
GPT-5.6 Sol Jul 2026 OpenAIreasoningweb								$5 / $30	1.05M
GPT-5.6 Terra Jul 2026 OpenAIreasoningweb	no published score							$2.50 / $15	1.05M
GPT-5.6 Luna Jul 2026 OpenAIreasoningweb								$1 / $6	1.05M
Gemini 3.1 Pro Feb 2026 Googlereasoningweb								$2 / $12	1M
Gemini 3.6 Flash Jul 2026 Googlereasoningweb	no published score		no published score		no published score			$1.50 / $7.50	1.05M
Gemini 3.5 Flash-Lite Jul 2026 Googlereasoningweb	no published score		no published score		no published score		no published score	$0.30 / $2.50	1.05M
Gemini 3.5 Flash May 2026 Googlereasoningweb							no published score	$1.50 / $9	1M
Grok 4.5 Jul 2026 xAIreasoningweb								$2 / $6	500K
Grok 4.1 Fast Nov 2025 xAIreasoningweb	no published score			no published score		no published score	no published score	$0.20 / $0.50	2M
Muse Spark 1.1 Jul 2026 Metareasoningweb							no published score	$1.25 / $4.25	1M
Kimi K3 Jul 2026 Moonshot AIreasoningweb		no published score						$3 / $15	1M
Inkling Jul 2026 Thinking Machinesreasoningopen							no published score	Free* / Free*	1M
Mistral Medium 3.5 Apr 2026 Mistral AIreasoningwebopen	no published score	no published score	no published score		no published score	no published score		Free* / Free*	256K
GLM-5.2 Jun 2026 Z.ai (GLM)reasoningopen	no published score						no published score	Free* / Free*	1M
DeepSeek V4 Pro Apr 2026 DeepSeekreasoningopen								Free* / Free*	—
Qwen 3.6 Apr 2026 Alibaba (Qwen)reasoningopen							no published score	Free* / Free*	—
Llama 4 Maverick Apr 2025 Metaopen	no published score	no published score		no published score	no published score		no published score	Free* / Free*	1M
Llama 4 Scout Apr 2025 Metaopen	no published score	no published score		no published score	no published score		no published score	Free* / Free*	10M

*Open-source models are free to download and run yourself; hosted-API pricing varies by vendor. Scores are provider-published evals where available, otherwise independent leaderboard runs, collected 2026-07-30. The dot next to each score shows who measured it: an independent run, the provider's own number, the provider's number where an independent run differs (hover for details). “reasoning” means the model thinks step by step before answering; “web” means the provider's assistant can search the live internet.