Model AtlasA selection map of agent foundation models 中文
Start researchModelsPapersGuideCompare 中文

Search the atlas

Esc to close · Cmd/Ctrl + K to open

Model families and generations

Expands by vendor → family → official generation name → checkpoint. The page does not assume every vendor has a continuous 1st, 2nd, 3rd generation.

Anthropic

Claude

Current catalog flag: Claude Fable 5, Claude Haiku 4.5, Claude Opus 4.8, Claude Sonnet 5

Claude 22023-07-11 · 1 models
Claude 2Other · Instruction-tuned
Claude 3.52024-06-20 · 1 models
Claude 3.5 SonnetOther · Vision
Claude 3.72025-02-19 · 1 models
Claude 3.7 SonnetOther · Reasoning
Claude 32024-03-04 · 1 models
Claude 3 OpusOther · Vision
Claude Fable 52026-06-09 · 1 models
Claude Fable 5Other · Vision · Current generation
Claude Haiku 4.52025-10-15 · 1 models
Claude Haiku 4.5Other · Vision · Current generation
Claude Mythos 52026-06-09 · 1 models
Claude Mythos 5Other · Vision
Claude Opus 42025-05-22 · 1 models
Claude Opus 4Other · Vision
Claude Opus 4.12025-08-05 · 1 models
Claude Opus 4.1Other · Vision
Claude Opus 4.52025-11-01 · 1 models
Claude Opus 4.5Other · Vision
Claude Opus 4.62026-02-05 · 1 models
Claude Opus 4.6Other · Vision
Claude Opus 4.72026-04-16 · 1 models
Claude Opus 4.7Other · Vision
Claude Opus 4.82026-05-28 · 1 models
Claude Opus 4.8Other · Vision · Current generation
Claude Sonnet 42025-05-22 · 1 models
Claude Sonnet 4Other · Vision
Claude Sonnet 4.52025-09-29 · 1 models
Claude Sonnet 4.5Other · Vision
Claude Sonnet 4.62026-02-17 · 1 models
Claude Sonnet 4.6Other · Vision
Claude Sonnet 52026-06-30 · 1 models
Claude Sonnet 5Other · Vision · Current generation

DeepSeek

DeepSeek-Coder

Current catalog flag: DeepSeek-Coder-V2-Lite-Instruct

V22024-06-27 · 1 models
DeepSeek-Coder-V2-Lite-InstructMixture of experts (MoE) · Coding · Current generation

DeepSeek

Current catalog flag: DeepSeek-V4-Flash, DeepSeek-V4-Pro

R12025-01-20 · 2 models
DeepSeek-R1Mixture of experts (MoE) · ReasoningDeepSeek-R1-0528Mixture of experts (MoE) · Reasoning
V32024-12-26 · 2 models
DeepSeek-V3Mixture of experts (MoE) · Instruction-tunedDeepSeek-V3-0324Mixture of experts (MoE) · Instruction-tuned
V3.12025-08-21 · 1 models
DeepSeek-V3.1Mixture of experts (MoE) · Instruction-tuned
V3.22025-12-01 · 3 models
DeepSeek-V3.2Mixture of experts (MoE) · Instruction-tunedDeepSeek-V3.2-ExpMixture of experts (MoE) · Instruction-tunedDeepSeek-V3.2-SpecialeMixture of experts (MoE) · Reasoning
V42026-04-24 · 2 models
DeepSeek-V4-FlashMixture of experts (MoE) · Instruction-tuned · Current generationDeepSeek-V4-ProMixture of experts (MoE) · Instruction-tuned · Current generation

Mistral AI

Mistral

Current catalog flag: Mistral Large 3

Devstral2025-07-10 · 2 models
Devstral MediumOther · Instruction-tunedDevstral SmallDense · Instruction-tuned
Magistral2025-06-10 · 2 models
Magistral MediumOther · ReasoningMagistral SmallDense · Reasoning
Mistral 7B (v0.3)2024-05-22 · 1 models
Mistral 7B Instruct v0.3Dense · Instruction-tuned
Large 32025-12-01 · 1 models
Mistral Large 3Mixture of experts (MoE) · Vision · Current generation
Medium 32025-05-07 · 1 models
Mistral Medium 3Other · Instruction-tuned
Mistral Small2025-01-28 · 1 models
Mistral Small 24B Instruct 2501Dense · Instruction-tuned
Small 42026-03-01 · 1 models
Mistral Small 4Mixture of experts (MoE) · Reasoning

Google

Gemini

Current catalog flag: Gemini 3.6 Flash

Gemini 1.52024-02-15 · 1 models
Gemini 1.5 ProMixture of experts (MoE) · Vision
Gemini 2.02025-02-05 · 1 models
Gemini 2.0 FlashOther · Vision
Gemini 2.52025-06-17 · 2 models
Gemini 2.5 FlashOther · VisionGemini 2.5 ProOther · Vision
Gemini 3.12026-02-19 · 1 models
Gemini 3.1 ProOther · Vision
Gemini 3.52026-05-19 · 1 models
Gemini 3.5 FlashOther · Vision
Gemini 3.62026-07-21 · 1 models
Gemini 3.6 FlashOther · Vision · Current generation
Gemini 32025-12-17 · 2 models
Gemini 3 FlashOther · VisionGemini 3 ProOther · Vision

Gemma

Current catalog flag: Gemma 4

Gemma 32025-03-12 · 3 models
Gemma 3 12B ITDense · VisionGemma 3 27BDense · VisionGemma 3 4B ITDense · Vision
Gemma 42026-04-02 · 1 models
Gemma 4Other · Vision · Current generation

GLM / Z.ai

GLM

Current catalog flag: GLM-5.2 · Older papers remain linked to the model used at publication

GLM-4.52025-07-28 · 1 models
GLM-4.5Mixture of experts (MoE) · Instruction-tuned
GLM-4.5 Air2025-07-28 · 1 models
GLM-4.5 AirMixture of experts (MoE) · Instruction-tuned
GLM-4.62025-09-30 · 1 models
GLM-4.6Mixture of experts (MoE) · Instruction-tuned
GLM-4.72025-12-22 · 1 models
GLM-4.7Mixture of experts (MoE) · Instruction-tuned
GLM-42024-06-18 · 1 models
GLM-4-9B-ChatDense · Instruction-tuned
GLM-52026-02-12 · 1 models
GLM-5Mixture of experts (MoE) · Instruction-tuned
GLM-5.12026-04-07 · 1 models
GLM-5.1Mixture of experts (MoE) · Instruction-tuned
GLM-5.22026-06-16 · 1 models
GLM-5.2Mixture of experts (MoE) · Instruction-tuned · Current generation

OpenAI

GPT

Current catalog flag: GPT-5.6 Luna, GPT-5.6 Sol, GPT-5.6 Terra

GPT-3.52023-03-01 · 2 models
GPT-3.5 TurboOther · Instruction-tunedtext-davinci-003Other · Instruction-tuned
GPT-42023-03-14 · 1 models
GPT-4Other · Instruction-tuned
GPT-4.12025-04-14 · 1 models
GPT-4.1Other · Instruction-tuned
GPT-4.52025-02-27 · 1 models
GPT-4.5Other · Instruction-tuned
GPT-4V2023-09-25 · 1 models
GPT-4VOther · Vision
GPT-4o2024-05-13 · 1 models
GPT-4oOther · Vision
GPT-52025-08-07 · 3 models
GPT-5Other · Instruction-tunedGPT-5 miniOther · Instruction-tunedGPT-5 nanoOther · Instruction-tuned
GPT-5.12025-11-13 · 1 models
GPT-5.1Other · Instruction-tuned
GPT-5.22025-12-11 · 1 models
GPT-5.2Other · Instruction-tuned
GPT-5.3 Codex2026-02-05 · 1 models
GPT-5.3 CodexOther · Coding
GPT-5.42026-03-05 · 1 models
GPT-5.4Other · Instruction-tuned
GPT-5.52026-04-23 · 1 models
GPT-5.5Other · Instruction-tuned
GPT-5.6 Luna2026-07-09 · 1 models
GPT-5.6 LunaOther · Instruction-tuned · Current generation
GPT-5.6 Sol2026-07-09 · 1 models
GPT-5.6 SolOther · Instruction-tuned · Current generation
GPT-5.6 Terra2026-07-09 · 1 models
GPT-5.6 TerraOther · Instruction-tuned · Current generation
gpt-oss2025-08-05 · 2 models
gpt-oss-120bMixture of experts (MoE) · Instruction-tunedgpt-oss-20bMixture of experts (MoE) · Instruction-tuned
o12024-12-17 · 1 models
o1Other · Reasoning
o32025-04-16 · 1 models
o3Other · Reasoning
o4-mini2025-04-16 · 1 models
o4-miniOther · Reasoning

xAI

Grok

Current catalog flag: Grok 4.5

Grok 32025-02-18 · 1 models
Grok 3Other · Instruction-tuned
Grok 42025-07-09 · 4 models
Grok 4Other · Instruction-tunedGrok 4.1Other · Instruction-tunedGrok 4.3Other · Instruction-tunedGrok 4.5Other · Instruction-tuned · Current generation

Moonshot AI

Kimi

Current catalog flag: Kimi K2 Thinking, Kimi K3

Kimi Audio2025-04-25 · 1 models
Kimi-Audio-7B-InstructOther · Instruction-tuned
Kimi Dev2025-06-17 · 1 models
Kimi-Dev-72BDense · Coding
K2.52026-01-27 · 1 models
Kimi K2.5Mixture of experts (MoE) · Instruction-tuned
K2.62026-04-20 · 1 models
Kimi K2.6Mixture of experts (MoE) · Instruction-tuned
K22025-07-11 · 2 models
Kimi K2 BaseMixture of experts (MoE) · BaseKimi K2 InstructMixture of experts (MoE) · Instruction-tuned
Kimi K22025-09-05 · 1 models
Kimi K2 Instruct 0905Mixture of experts (MoE) · Instruction-tuned
K2 Thinking2025-11-06 · 1 models
Kimi K2 ThinkingMixture of experts (MoE) · Reasoning · Current generation
K32026-07-16 · 1 models
Kimi K3Mixture of experts (MoE) · Instruction-tuned · Current generation
Kimi Linear2025-12-05 · 1 models
Kimi Linear 48B-A3B InstructMixture of experts (MoE) · Instruction-tuned
Kimi VL2025-06-10 · 1 models
Kimi-VL-A3B-Thinking-2506Mixture of experts (MoE) · Reasoning
Moonlight2025-02-23 · 1 models
Moonlight-16B-A3B-InstructMixture of experts (MoE) · Instruction-tuned

Meta

Llama

Current catalog flag: Llama 4 Maverick 17B-128E

Llama 22023-07-18 · 2 models
Llama 2 13B ChatDense · Instruction-tunedLlama 2 70B ChatDense · Instruction-tuned
Llama 3.12024-07-23 · 3 models
Llama 3.1 405B InstructDense · Instruction-tunedLlama 3.1 70B InstructDense · Instruction-tunedLlama 3.1 8B InstructDense · Instruction-tuned
Llama 3.22024-09-25 · 1 models
Llama 3.2 11B Vision InstructDense · Vision
Llama 3.32024-12-06 · 1 models
Llama 3.3 70B InstructDense · Instruction-tuned
Llama 32024-04-18 · 2 models
Llama 3 70B InstructDense · Instruction-tunedLlama 3 8B InstructDense · Instruction-tuned
Llama 42025-04-05 · 2 models
Llama 4 Maverick 17B-128EMixture of experts (MoE) · Vision · Current generationLlama 4 Scout 17B-16EMixture of experts (MoE) · Vision

Muse

Current catalog flag: Muse Spark

Muse Spark2026-04-08 · 1 models
Muse SparkOther · Reasoning · Current generation

MiniMax

MiniMax

Current catalog flag: MiniMax M3

M12025-06-16 · 1 models
MiniMax M1Mixture of experts (MoE) · Instruction-tuned
M22025-10-27 · 2 models
MiniMax M2Mixture of experts (MoE) · Instruction-tunedMiniMax M2.7Mixture of experts (MoE) · Instruction-tuned
M32026-06-01 · 1 models
MiniMax M3Mixture of experts (MoE) · Instruction-tuned · Current generation

Qwen / Alibaba

Qwen

Current catalog flag: Qwen3.6-35B-A3B

Qwen 12023-09-25 · 1 models
Qwen 7B ChatDense · Instruction-tuned
Qwen2.52024-09-19 · 14 models
Qwen2.5-0.5BDense · BaseQwen2.5-0.5B-InstructDense · Instruction-tunedQwen2.5-1.5BDense · BaseQwen2.5-1.5B-InstructDense · Instruction-tunedQwen2.5-14BDense · BaseQwen2.5-14B-InstructDense · Instruction-tunedQwen2.5-32BDense · BaseQwen2.5-32B-InstructDense · Instruction-tunedQwen2.5-3BDense · BaseQwen2.5-3B-InstructDense · Instruction-tunedQwen2.5-72BDense · BaseQwen2.5-72B-InstructDense · Instruction-tunedQwen2.5-7BDense · BaseQwen2.5-7B-InstructDense · Instruction-tuned
Qwen22024-06-06 · 1 models
Qwen2-72B-InstructDense · Instruction-tuned
Qwen32025-04-29 · 6 models
Qwen3-1.7BDense · ReasoningQwen3-235B-A22BMixture of experts (MoE) · Instruction-tunedQwen3-30B-A3BMixture of experts (MoE) · Instruction-tunedQwen3-32BDense · Instruction-tunedQwen3-8BDense · ReasoningQwen3-MaxMixture of experts (MoE) · Instruction-tuned
Qwen3.52026-04-24 · 1 models
Qwen3.5-27BDense · Vision
Qwen3.62026-04-16 · 1 models
Qwen3.6-35B-A3BMixture of experts (MoE) · Vision · Current generation
Qwen3-Coder2025-07-22 · 1 models
Qwen3-Coder-480B-A35B-InstructMixture of experts (MoE) · Coding
Qwen3-Next2025-04-29 · 1 models
Qwen3-Next-80B-A3B-InstructMixture of experts (MoE) · Instruction-tuned
Qwen3-VL2025-04-29 · 1 models
Qwen3-VL-235B-A22B-InstructMixture of experts (MoE) · Instruction-tuned

LMSYS

Vicuna

Current catalog flag: Vicuna 13B · Older papers remain linked to the model used at publication

Vicuna v12023-03-30 · 1 models
Vicuna 13BDense · Instruction-tuned · Current generation