Providers

1OpenAI70.6

Best model

GPT-6 Astra

Ranked

38 · 4 open

From

$0.085/M

Maker of the GPT and o-series models and ChatGPT. Sells API access, ChatGPT subscriptions and enterprise deals; deeply tied to Microsoft Azure.

2Anthropic70.4

Best model

Claude Fable 5

Ranked

20

From

$0.500/M

Safety-focused lab behind the Claude family (Haiku, Sonnet, Opus and the Claude 5 tier). Strong in coding and agentic work; sells API access, Claude apps and Claude Code.

3Moonshot AI66.5

Best model

Kimi K3

Ranked

8 · 7 open

From

$1.07/M

Maker of the Kimi assistant and the open-weights Kimi K2 / K3 mixture-of-experts models with strong agentic scores.

4Meta65.4

Best model

Muse Spark

Ranked

21 · 14 open

From

$0.053/M

Ships the open-weights Llama family and Meta AI products; announced the Superintelligence Labs reorganisation in 2025 and very large data-centre investments.

5Google64.4

Best model

Gemini 3.7 Flash

Ranked

34 · 9 open

From

$0.040/M

Google DeepMind builds the Gemini models (Pro, Flash, Flash-Lite, Deep Think) and the open Gemma family; distributed through Gemini apps, Vertex AI and AI Studio.

6Alibaba63.9

Best model

Qwen3 8

Ranked

44 · 44 open

From

$0.030/M

Alibaba Cloud's Qwen team ships the broadest open-weights line-up (Qwen3, Qwen3.5, coder and VL variants) plus hosted Qwen Max models.

7xAI63.8

Best model

Grok 4.6

Ranked

14

From

$1.25/M

Elon Musk's lab behind Grok, trained on the Colossus cluster in Memphis; distributes through X and an API.

8Zhipu AI61.0

Best model

GLM-5.3-Flash

Ranked

14 · 14 open

From

$0.119/M

Tsinghua spin-out behind the GLM series, released open-weights (GLM-4.5, 4.6, 5) with a focus on agentic coding.

Best model

Mimo v2 Pro

Ranked

5 · 3 open

From

$0.150/M

MiMo models from Xiaomi's AI lab, aimed at on-device and assistant use.

10DeepSeek60.3

Best model

Deepseek v4 Pro

Ranked

14 · 10 open

From

$0.068/M

Open-weights lab known for cost-efficient training (V3, R1, V4) and aggressive API pricing.

11NVIDIA58.3

Best model

Nemotron 3 Ultra 550B A55B

Ranked

19 · 32 open

From

$0.063/M

The GPU maker also publishes open Nemotron models fine-tuned from Llama and its own architectures, mostly as showcases for its inference stack.

Best model

KAT-Coder-Pro V2

Ranked

1

From

$0.525/M

13MiniMax56.9

Best model

MiniMax-M3

Ranked

5 · 7 open

From

$0.525/M

Consumer-app-driven lab (Talkie, Hailuo video) that also ships the open-weights MiniMax M-series text models.

14Upstage56.0

Best model

Solar Pro 4

Ranked

3 · 2 open

From

$0.262/M

Best model

Inkling-Small

Ranked

1 · 1 open

From

$0.638/M

16StepFun54.9

Best model

Step 3.5 Flash

Ranked

4 · 3 open

From

$0.150/M

Step series multimodal and reasoning models.

Best model

Ling 3.0 Flash

Ranked

1 · 2 open

From

$0.115/M

18Cohere53.9

Best model

Command A+

Ranked

7 · 13 open

From

$4.38/M

Enterprise-focused lab behind Command A and the Aya multilingual models, with a private-deployment emphasis.

19Amazon52.3

Best model

Nova 2.0 Pro Preview

Ranked

7

From

$0.061/M

AWS's Nova family sits alongside Bedrock's third-party catalogue; Amazon is also Anthropic's largest investor.

20ByteDance52.3

Best model

Doubao-Seed-Code

Ranked

1

From

–

ByteDance Seed builds the Doubao/Seed models powering the Doubao assistant, one of China's most used chatbots.

Best model

JT-35B-Flash

Ranked

2

From

–

22IBM51.3

Best model

Granite 4.2 30B

Ranked

12 · 10 open

From

$0.053/M

Granite open models aimed at enterprise deployment through watsonx.

Best model

Ring-2.6-1T

Ranked

1

From

$0.850/M

Best model

K-EXAONE

Ranked

3 · 5 open

From

–

25Mistral AI50.8

Best model

Mistral Medium 3.5

Ranked

24 · 28 open

From

$0.090/M

Europe's leading lab, mixing open-weights (Mistral Small, Devstral) and commercial models (Medium, Large, Magistral); strong enterprise and sovereign-cloud positioning.

26Inception49.5

Best model

Mercury 2

Ranked

1

From

$0.375/M

Diffusion-LLM pioneer; Mercury models generate tokens in parallel for very high output speed.

Best model

HyperNova 60B 2605 (high, based on gpt-oss-120b)

Ranked

1 · 1 open

From

$0.065/M

Best model

K2 Think V2

Ranked

1 · 2 open

From

–

Best model

Trinity Large

Ranked

1

From

$0.413/M

30Baidu48.4

Best model

Ernie 5.0

Ranked

2 · 2 open

From

$0.485/M

ERNIE models (4.5 open-weights, 5 series) behind the ERNIE Bot assistant and Baidu Cloud.

Best model

Kat Coder Pro v1

Ranked

3 · 3 open

From

–

Best model

Intellect 3

Ranked

1

From

–

Best model

Ring Flash 2.0

Ranked

2

From

$0.248/M

Best model

MiniCPM5-1B

Ranked

2 · 3 open

From

–

Best model

Olmo 3.1 32B

Ranked

3

From

–

36Liquid AI44.3

Best model

LFM2.5-8B-A1B

Ranked

5 · 7 open

From

$0.053/M

Best model

Hermes 4 - Llama-3.1 405B

Ranked

2 · 5 open

From

$1.50/M

Best model

Tri-21B-Think

Ranked

1 · 1 open

From

–

Best model

LongCat Flash Lite

Ranked

1 · 2 open

From

–

Best model

Motif-2-12.7B

Ranked

1

From

–

Best model

HyperCLOVA X SEED Think (32B)

Ranked

1 · 1 open

From

–

Best model

Falcon-H1R-7B

Ranked

1 · 1 open

From

–

Best model

Reka Flash

Ranked

1

From

$0.350/M

Best model

Mi:dm K 2.5 Pro

Ranked

1

From

–

45AI21 Labs42.7

Best model

Jamba 1.5 Large

Ranked

7 · 3 open

From

$0.250/M

Best model

Nanbeige4.1-3B

Ranked

1 · 1 open

From

–

Best model

Sarvam 105B (high)

Ranked

2 · 2 open

From

$0.050/M

Best model

Yi-1.5-34B

Ranked

2

From

–

Best model

Olmo 3 7B Think

Ranked

2 · 2 open

From

$0.125/M

50Microsoft39.7

Best model

Phi-4

Ranked

3 · 3 open

From

$0.219/M

Publishes the small open Phi models and MAI in-house models alongside its OpenAI partnership and Azure AI Foundry distribution.

Best model

Reka Flash 3

Ranked

1 · 1 open

From

$0.350/M

Best model

Apertus 70B Instruct

Ranked

2 · 2 open

From

$0.125/M

53Poolside37.5

Best model

Laguna M.1

Ranked

2 · 1 open

From

–

Best model

Qwen3 0 6B

Ranked

1 · 2 open

From

–