Benchmark Almanac

Labs

We read every model card from the first 51 labs below. After them comes every other lab with a scored model. Open a lab to see its models over time and the benchmarks its cards report.

OpenAI

Best model
GPT-6 Astra, 169.1
Latest
GPT-6.1 Terra, Oct 2026
Tracked
90 models, 48 model cards

Anthropic

Best model
Claude Opus 5.5, 168.4
Latest
Claude Haiku 5.5, Oct 2026
Tracked
29 models, 23 model cards

Google DeepMind

Best model
Gemini 4 Argon, 166.1
Latest
Gemini 4 Carbon, Oct 2026
Tracked
76 models, 44 model cards

Meta

Best model
Muse Wonder, 161.7
Latest
Muse Spark Open, Oct 2026
Tracked
39 models, 17 model cards

DeepSeek

Best model
DeepSeek V4.1 Flash 1017, 158.8
Latest
DeepSeek V4.1 Flash 1017, Oct 2026
Tracked
38 models, 20 model cards

Moonshot AI

Best model
Kimi K3, 158.3
Latest
Kimi K3.1 Pro, Oct 2026
Tracked
15 models, 14 model cards

Qwen

Best model
Qwen 3.8 Max, 158.0
Latest
Qwen3.8-Omni-Flash, Sep 2026
Tracked
94 models, 32 model cards

xAI

Best model
Grok 4.6, 156.9
Latest
Grok Voice 3.0, Oct 2026
Tracked
25 models, 17 model cards

Z.ai

Best model
GLM-5.3, 156.2
Latest
GLM Legal, Oct 2026
Tracked
27 models, 16 model cards

Mistral AI

Best model
Mistral Large 4, 150.6
Latest
Mistral Large 4 (Le Chonk, open weights), Oct 2026
Tracked
47 models, 30 model cards

Thinking Machines

Best model
Inkling-Small, 150.5
Latest
Inkling, Jul 2026
Tracked
2 models, 2 model cards

Reflection

Best model
Beam 1.0, 150.2
Latest
Beam 1.0, Oct 2026
Tracked
1 model, 1 model card

MiniMax

Best model
MiniMax-M3, 146.9
Latest
MiniMax-M3.1-Flash, Oct 2026
Tracked
9 models, 8 model cards

NVIDIA

Best model
Nemotron 3 Ultra, 146.0
Latest
Nemotron 3.5 Bolt E4B, Oct 2026
Tracked
29 models, 20 model cards

Microsoft

Best model
Phi-4, 130.6
Latest
Mage-VL-Cowork, Oct 2026
Tracked
28 models, 18 model cards

Cohere

Best model
Command R+, 119.0
Latest
North Small Translate 1.0, Sep 2026
Tracked
15 models, 10 model cards

Technology Innovation Institute

Best model
Falcon-180B, 112.0
Latest
Falcon-Emirati-7B, Oct 2026
Tracked
11 models, 4 model cards

Baichuan

Best model
Baichuan2-13B, 103.1
Latest
Baichuan-M3-235B, Jan 2026
Tracked
6 models, 2 model cards

Prime Intellect

Best model
INTELLECT-1, 100.8
Latest
INTELLECT-3, Nov 2025
Tracked
3 models, 2 model cards

Shanghai AI Lab

Best model
Not on the index
Latest
Intern-S2-397B, Sep 2026
Tracked
44 models, 5 model cards

ByteDance Seed

Best model
Not on the index
Latest
Seed2.1 Pro, Jun 2026
Tracked
33 models, 15 model cards

Ai2

Best model
Not on the index
Latest
MolmoAct 2, May 2026
Tracked
16 models, 8 model cards

Xiaomi Corp

Best model
Not on the index
Latest
MiMo-V2.6 Flash, Sep 2026
Tracked
12 models, 8 model cards

IBM

Best model
Not on the index
Latest
Granite 4.2 30B, Aug 2026
Tracked
11 models, 6 model cards

Liquid AI

Best model
Not on the index
Latest
LFM3-1.7B, Oct 2026
Tracked
10 models, 10 model cards

Tencent

Best model
Not on the index
Latest
Tencent Hy4 preview, Aug 2026
Tracked
9 models, 8 model cards

Amazon

Best model
Not on the index
Latest
Nova 2.0 Pro Preview, Dec 2025
Tracked
8 models, 5 model cards

Upstage

Best model
Not on the index
Latest
Solar Mini 4, Oct 2026
Tracked
8 models, 7 model cards

Ant Group

Best model
Not on the index
Latest
Ling-3.1-flash, Oct 2026
Tracked
7 models, 5 model cards

Kuaishou

Best model
Not on the index
Latest
KAT-Coder-V2.5-Dev, Jul 2026
Tracked
7 models, 4 model cards

Nous Research

Best model
Not on the index
Latest
NousCoder-14B, Jan 2026
Tracked
6 models, 4 model cards

Baidu

Best model
Not on the index
Latest
ERNIE 5.1, May 2026
Tracked
6 models, 6 model cards

StepFun

Best model
Not on the index
Latest
Step 5 Preview, Sep 2026
Tracked
5 models, 5 model cards

Reka

Best model
Not on the index
Latest
Reka Edge, Mar 2026
Tracked
5 models, 3 model cards

Cognition

Best model
Not on the index
Latest
SWE-2, Sep 2026
Tracked
5 models, 5 model cards

Inception Labs

Best model
Not on the index
Latest
Mercury 3, Oct 2026
Tracked
5 models, 5 model cards

Meituan

Best model
Not on the index
Latest
LongCat-2.0, Jul 2026
Tracked
5 models, 5 model cards

LG AI Research

Best model
Not on the index
Latest
K-EXAONE 2.0, Jul 2026
Tracked
5 models, 5 model cards

Arcee AI

Best model
Not on the index
Latest
Trinity Large Thinking, Apr 2026
Tracked
4 models, 4 model cards

Cursor

Best model
Not on the index
Latest
Composer 2.5, May 2026
Tracked
4 models, 4 model cards

Poolside

Best model
Not on the index
Latest
Laguna S 2.1, Jul 2026
Tracked
4 models, 4 model cards

Deep Cogito

Best model
Not on the index
Latest
Cogito v2.1, Nov 2025
Tracked
3 models, 3 model cards

Sarvam AI

Best model
Not on the index
Latest
Sarvam 105B, Mar 2026
Tracked
3 models, 3 model cards

Physical Intelligence

Best model
Not on the index
Latest
π0.5, Apr 2025
Tracked
3 models, 3 model cards

Perplexity

Best model
Not on the index
Latest
sonar, Jan 2025
Tracked
2 models, 2 model cards

Apple

Best model
Not on the index
Latest
AFM 3 Cloud, Jun 2026
Tracked
2 models, 2 model cards

Stanford University

Best model
Not on the index
Latest
OpenVLA-OFT, Feb 2025
Tracked
2 models, 2 model cards

Toyota Research Institute

Best model
Not on the index
Latest
Foundry-Qwen3VLA-2.1B, Apr 2026
Tracked
2 models, 2 model cards

iFlytek

Best model
Not on the index
Latest
Spark-X2.6-4B, Oct 2026
Tracked
1 model, 1 model card

SenseTime

Best model
Not on the index
Latest
SenseNova-SI-1.3-InternVL3-8B, Jan 2026
Tracked
1 model, 1 model card

Swiss AI Initiative

Best model
Not on the index
Latest
Apertus 70B, Sep 2025
Tracked
1 model, 1 model card

Other labs with scored models