AI model directory

Which AI model fits your task?

This directory covers the models in our current comparisons. Open a profile for specifications, published measurements, pricing, and relevant matchups.

Models: 87
LLMs: 36
Image models: 28
Video models: 23

Browse the catalog

Each profile brings together technical facts, sources, and relevant direct comparisons.

87 of 87 models

LLMs

36 models

Language modelActive

Claude Fable 5.1

Released September 1, 2026. Anthropic positions Fable 5.1 for long-running agents, advanced coding, multistep research, and document, spreadsheet, and presentation work. 1M context, 128K output, adaptive thinking, and high effort by default. Official API ID: claude-fable-5-1.

Released: September 2026

Model type: Proprietary
Context window: 1,000,000 tokens
Knowledge cutoff: June 2026

Open model profile

Language modelLimited access

Claude Mythos 5.1

Released September 1, 2026. Shares the model, specifications, and pricing of Fable 5.1, but uses fewer restrictive safeguards. Available by invitation only through Project Glasswing. Official API ID: claude-mythos-5-1.

Released: September 2026

Model type: Proprietary
Context window: 1,000,000 tokens
Knowledge cutoff: June 2026

Open model profile

Language modelActive

Claude Opus 5

Released July 24, 2026. Near-frontier intelligence at launch, close to Fable 5 at half the price ($5/$25). 1M context, 128K output, and a May 2026 knowledge cutoff. It became the new default on Claude Max and strongest model on Claude Pro at launch. It led Frontier-Bench v0.1, surpassed Fable 5 on OSWorld 2.0, came within 0.5% of Fable 5 on CursorBench 3.2, and trailed Mythos 5 on cybersecurity. Those results describe the pre-Fable-5.1 and pre-Mythos-5.1 cohort.

Released: July 2026

Model type: Proprietary
Context window: 1,000,000 tokens
Knowledge cutoff: May 2026

Open model profile

Language modelActive

Claude Fable 5

Mythos-class tier above Opus, most capable widely-released model, 1M context (new tokenizer, ~30% more tokens per text vs. pre-Opus-4.7 models), 128K output, adaptive thinking only, safety classifiers with refusal fallback to Opus 4.8. Briefly suspended 2026-06-12 to 2026-06-30 under a US export control directive after a jailbreak was found; generally available again since 2026-07-01.

Released: June 2026

Model type: Proprietary
Context window: 1,000,000 tokens
Knowledge cutoff: January 2026

Open model profile

Language modelActive

Claude Sonnet 5

Released 2026-06-30, best speed/intelligence balance and the default model for Claude Free and Pro plans, rivals Opus 4.8 at lower cost. 1M context, 128K output, adaptive thinking (effort defaults to high on the API and Claude Code), uses the Opus 4.7 tokenizer (~30% more tokens per text). API pricing is $2/$10 per 1M tokens. Knowledge cutoff January 2026.

Released: June 2026

Model type: Proprietary
Context window: 1,000,000 tokens
Knowledge cutoff: January 2026

Open model profile

Language modelLimited access

Claude Mythos 5

Shares Fable 5's specs and pricing, without the safety classifiers. Invitation-only limited availability via Project Glasswing for vetted cybersecurity/biomedical/infrastructure partners, no self-serve signup. Briefly suspended 2026-06-12 to 2026-06-30 under the same export control directive as Fable 5; access restored 2026-07-01.

Released: June 2026

Model type: Proprietary
Context window: 1,000,000 tokens
Knowledge cutoff: January 2026

Open model profile

Language modelLimited access

GPT-6 Astra

New OpenAI flagship with a 1.05 million token context window, 128,000 maximum output tokens, and an April 30, 2026 knowledge cutoff. Access has been rolling out in phases since September 3, 2026. OpenAI has not announced smaller GPT-6 variants.

Released: September 2026

Model type: Proprietary
Context window: 1,050,000 tokens
Knowledge cutoff: April 30, 2026

Open model profile

Language modelActive

GPT-5.6 Sol

Flagship of the GPT-5.6 family (Sol, Terra, Luna). Previewed 2026-06-26, generally available from 2026-07-09 after the US government lifted the restricted-access framework. Available in ChatGPT (Plus, Pro, Business, Enterprise from medium effort setting; Pro and Enterprise also get Sol Pro), Codex, and the API. Terminal-Bench 2.1 88.8% (Sol Ultra mode with four parallel sub-agents: 91.9%), Artificial Analysis Coding Agent Index 80, SWE-Bench Pro 64.6% (about 15 points behind Claude Mythos 5 and Fable 5). New at GA: programmatic tool calling with model-written JavaScript in a V8 sandbox.

Released: June 2026

Model type: Proprietary
Context window: 1,050,000 tokens
Knowledge cutoff: February 16, 2026

Open model profile

Language modelActive

GPT-5.6 Terra

Balanced tier of the GPT-5.6 family. Generally available from 2026-07-09. Default model for Free and Go users on ChatGPT Work and Codex, effort-selectable on paid tiers. Matches GPT-5.5 quality at roughly half the price ($2.50 input / $15 output per 1M tokens). Terminal-Bench 2.1 82.5%.

Released: June 2026

Model type: Proprietary
Context window: 1,050,000 tokens
Knowledge cutoff: February 16, 2026

Open model profile

Language modelActive

GPT-5.6 Luna

Fast and cheapest tier of the GPT-5.6 family. Generally available from 2026-07-09. Effort-selectable on ChatGPT Work and Codex for Plus, Pro, Business, and Enterprise users. $1 input / $6 output per 1M tokens. Terminal-Bench 2.1 84.3%.

Released: June 2026

Model type: Proprietary
Context window: 1,050,000 tokens
Knowledge cutoff: February 16, 2026

Open model profile

Language modelActive

GPT-5.5

1M context, agentic workflows, Terminal-Bench 2.0 82.7%, parameter count not disclosed

Released: April 2026

Model type: Proprietary
Context window: 1,050,000 tokens
Knowledge cutoff: December 1, 2025

Open model profile

Language modelActive

GPT-5.4

1M context, parameter count not disclosed

Released: March 2026

Model type: Proprietary
Context window: 1,050,000 tokens
Knowledge cutoff: August 31, 2025

Open model profile

Language modelActive

GPT-5.4 mini

400K context, parameter count not disclosed

Released: March 2026

Model type: Proprietary
Context window: 400,000 tokens
Knowledge cutoff: August 31, 2025

Open model profile

Language modelActive

Gemini 3.8 Flash

Released September 2, 2026. Multimodal Flash model with a 1,048,576-token context window, up to 65,536 output tokens, and three reasoning levels. Google does not publish its parameter count or architecture.

Released: September 2026

Model type: Proprietary
Context window: 1,048,576 tokens
Knowledge cutoff: March 2026

Open model profile

Language modelActive

Gemini 3.7 Flash

Released August 13, 2026. Multimodal Flash model with a 1,048,576-token context window, up to 65,536 output tokens, and three reasoning levels. Google does not publish its parameter count or architecture.

Released: August 2026

Model type: Proprietary
Context window: 1,048,576 tokens
Knowledge cutoff: March 2026

Open model profile

Language modelActive

Gemini 3.6 Flash

Released July 21, 2026. Google does not publish its parameter count, architecture, or knowledge cutoff.

Released: July 2026

Model type: Proprietary
Context window: 1,048,576 tokens

Open model profile

Language modelActive

Gemini 3.5 Flash

Released May 19, 2026. Google does not publish its parameter count, architecture, or knowledge cutoff.

Released: May 2026

Model type: Proprietary
Context window: 1,048,576 tokens

Open model profile

Language modelPreview

Gemini 3.1 Pro Preview

2x reasoning improvement over Gemini 3 Pro, GPQA Diamond 94.3%, SWE-bench 80.6%

Released: February 2026

Model type: Proprietary
Context window: 1,048,576 tokens

Open model profile

Language modelActive

Gemini 3.5 Flash-Lite

Announced July 21, 2026, high-throughput/low-latency Lite model for agentic search and document processing, Terminal-Bench 2.1 54%, SWE-Bench Pro 54.2%, OSWorld 74%

Released: July 2026

Model type: Proprietary
Context window: 1,048,576 tokens

Open model profile

Language modelActive

Muse Spark 1.3

Multimodal reasoning model for long-running agentic and coding tasks. Meta reports about 20% fewer tool calls and about 25% fewer tokens than Muse Spark 1.2; parameter count is not published.

Released: September 2026

Model type: Proprietary
Context window: 1,048,576 tokens

Open model profile

Language modelActive

Grok 4.6

Flagship for coding, knowledge work, and long-running agents with a 500,000-token context window. Supports text and image input, four reasoning levels, and integrated tools. Parameter count not disclosed.

Released: August 2026

Model type: Proprietary
Context window: 500,000 tokens
Knowledge cutoff: February 1, 2026

Open model profile

Language modelActive

Grok 4.5

Coding/agentic flagship (first release under SpaceXAI), 500K context, parameter count not disclosed

Released: July 2026

Model type: Proprietary
Context window: 500,000 tokens
Knowledge cutoff: February 1, 2026

Open model profile

Language modelActive

Grok 4.3

1M context, parameter count not disclosed

Released: May 2026

Model type: Proprietary
Context window: 1,000,000 tokens

Open model profile

Language modelPreview

DeepSeek-V4-Pro

1.6T total, 49B active, 1M context, CSA+HCA attention, MIT-licensed

Released: April 2026

Model type: Open weights
Parameters: 1.6T
Active parameters: 49B
Context window: 1,000,000 tokens

Open model profile

Language modelPreview

DeepSeek-V4-Flash

284B total, 13B active, 1M context, MIT-licensed

Released: April 2026

Model type: Open weights
Parameters: 284B
Active parameters: 13B
Context window: 1,000,000 tokens

Open model profile

Language modelActive

Kimi K3

2.8T total, 104B active, 896 experts with 16 active experts per token, KDA hybrid linear attention, native vision, and a 1,048,576-token context window. Model weights are published.

Released: July 2026

Model type: Open weights
Parameters: 2.8T
Active parameters: 104B
Context window: 1,048,576 tokens

Open model profile

Language modelActive

Kimi K2.6

1T total, 32B active, 384 experts (8 routed + 1 shared), 262K context, native INT4, Modified MIT

Released: April 2026

Model type: Open weights
Parameters: 1T
Active parameters: 32B
Context window: 262,144 tokens

Open model profile

Language modelOpen source

GLM-5.3-Flash

Natively multimodal open-weights MoE with 320B total parameters, 18B active parameters, and a 1M-token context window.

Released: August 2026

Model type: Open weights
Parameters: 320B
Active parameters: 18B
Context window: 1,000,000 tokens

Open model profile

Language modelActive

GLM-5.3

Released on August 14, 2026. Same base as GLM-5.2 with 744B total and 40B active parameters. The weights are available from Z.ai on Hugging Face.

Released: August 2026

Model type: Open weights
Parameters: 744B
Active parameters: 40B
Context window: 1,000,000 tokens

Open model profile

Language modelOpen source

GLM-5.2

Released 2026-06-16, 744B total parameters, 40B active, 1M context, MIT-licensed weights

Released: June 2026

Model type: Open weights
Parameters: 744B
Active parameters: 40B
Context window: 1,000,000 tokens

Open model profile

Language modelOpen source

GLM-5.1

MIT-licensed model with 744B total parameters, 40B active, and a 200K context window.

Released: April 2026

Model type: Open weights
Parameters: 744B
Active parameters: 40B
Context window: 200,000 tokens

Open model profile

Language modelAPI only

Qwen 3.8 Max 0902

The Qwen3.8-Max-0902 snapshot released September 2, 2026 improves coding, long-horizon autonomous development, agent collaboration, and visual analysis. The MoE model has 2.4T total parameters, 95B active parameters, and a 1M-token context window.

Released: September 2026

Model type: Proprietary
Parameters: 2.4T
Active parameters: 95B
Context window: 1,000,000 tokens

Open model profile

Language modelActive

Mistral Large 3

Sparse MoE with 675B total, 41B active

Released: December 2025

Model type: Open weights
Parameters: 675B
Active parameters: 41B
Context window: 256,000 tokens

Open model profile

Language modelActive

Mistral Medium 3.5

Coding-focused current API model; parameter count not disclosed

Released: April 2026

Model type: Open weights
Context window: 256,000 tokens

Open model profile

Language modelOpen source

MiMo-V2.5

MIT-licensed omni-modal model with a 1M context window; parameter count not disclosed on the official model page.

Released: April 2026

Model type: Open weights
Context window: 1,000,000 tokens

Open model profile

Language modelOpen source

MiMo-V2.5-Pro

1T total parameters, 42B active parameters, and a 1M context window.

Released: April 2026

Model type: Open weights
Parameters: 1T
Active parameters: 42B
Context window: 1,000,000 tokens

Open model profile

Image models

28 models

GPT Image 2

Released: April 2026

Weights: Proprietary
Access: API, Web
Maximum output: 4K
Benchmark configuration: High

Open model profile

Reve 2.1

Released: July 2026

Weights: Proprietary
Access: API, Web
Maximum output: 4K

Open model profile

Image modelPreview

Nano Banana 2

Released: February 2026

Weights: Proprietary
Access: API, Web
Maximum output: 4K
Benchmark configuration: Gemini 3.1 Flash Image Preview

Open model profile

Image modelDeprecated

GPT Image 1.5

Released: December 2025

Weights: Proprietary
Access: API, Web
Maximum output: 1,536 × 1,024 pixels
Benchmark configuration: High

Open model profile

Image modelPreview

MAI-Image-2.5

Released: June 2026

Weights: Proprietary
Access: API, Web
Parameters: 20B
Maximum output: 1,048,576 pixels

Open model profile

Nano Banana Pro

Released: November 2025

Weights: Proprietary
Access: API, Web
Maximum output: 4K
Benchmark configuration: Gemini 3 Pro Image

Open model profile

Nano Banana 2 Lite

Released: June 2026

Weights: Proprietary
Access: API, Web
Maximum output: 1K
Benchmark configuration: Gemini 3.1 Flash Lite Image

Open model profile

Seedream 5.0 Pro

Released: July 2026

Weights: Proprietary
Access: API, Web
Maximum output: 2K

Open model profile

Grok Imagine Image Quality

Released: April 2026

Weights: Proprietary
Access: API, Web
Maximum output: 2K
Benchmark configuration: Quality

Open model profile

Black Forest Labs

FLUX.2 Max

Released: December 2025

Weights: Proprietary
Access: API, Web
Maximum output: 4 megapixels
Benchmark configuration: Max

Open model profile

Qwen Image 2.0 Pro

Released: April 2026

Weights: Proprietary
Access: API
Maximum output: 2,048 × 2,048 pixels
Benchmark configuration: 2026-04-22

Open model profile

Image modelPreview

MAI-Image-2.5 Flash

Released: June 2026

Weights: Proprietary
Access: API, Web
Maximum output: 1,048,576 pixels
Benchmark configuration: Flash

Open model profile

HiDream-O1-Image-1.5

Released: June 2026

Weights: Open weights
Access: API, Web, Local
Parameters: 8B
Maximum output: 2,048 × 2,048 pixels

Open model profile

Krea 2 Medium Turbo

Released: June 2026

Weights: Proprietary
Access: API, Web
Maximum output: 1K
Benchmark configuration: Medium Turbo

Open model profile

Seedream 4.0

Released: September 2025

Weights: Proprietary
Access: API, Web
Maximum output: 4K

Open model profile

Krea 2 Large

Released: May 2026

Weights: Proprietary
Access: API, Web
Maximum output: 1K
Benchmark configuration: Large

Open model profile

Luma UNI 1 Max

Released: May 2026

Weights: Proprietary
Access: API, Web
Maximum output: 2K
Benchmark configuration: Max

Open model profile

Black Forest Labs

FLUX.2 Flex

Released: November 2025

Weights: Proprietary
Access: API, Web
Maximum output: 4 megapixels
Benchmark configuration: Flex

Open model profile

Recraft V4.1 Utility

Released: May 2026

Weights: Proprietary
Access: API, Web
Maximum output: 1 megapixels
Benchmark configuration: Utility

Open model profile

Ideogram 4.0

Released: June 2026

Weights: Open weights
Access: API, Web, Local
Parameters: 9.3B
Maximum output: 2K

Open model profile

Wan 2.6 Text to Image

Released: January 2026

Weights: Proprietary
Access: API
Maximum output: 2K
Benchmark configuration: Text to Image

Open model profile

Black Forest Labs

FLUX.2 Pro

Released: November 2025

Weights: Proprietary
Access: API, Web
Maximum output: 4 megapixels
Benchmark configuration: Pro

Open model profile

HunyuanImage 3.0 Instruct

Released: January 2026

Weights: Open weights
Access: API, Local
Parameters: 80B
Active parameters: 13B

Open model profile

Black Forest Labs

FLUX.2 Klein 9B

Released: January 2026

Weights: Open weights
Access: API, Local
Parameters: 9B
Maximum output: 4 megapixels

Open model profile

HiDream-O1-Image

Released: May 2026

Weights: Open weights
Access: API, Local
Parameters: 8B
Maximum output: 2,048 × 2,048 pixels

Open model profile

Firefly Image Model 5

Released: October 2025

Weights: Proprietary
Access: API, Web
Maximum output: 4 megapixels
Status: Generally available since March 19, 2026

Open model profile

Midjourney V8.1

Released: April 2026

Weights: Proprietary
Access: Web
Maximum output: 2K

Open model profile

Stable Diffusion 3.5 Large

Released: October 2024

Weights: Open weights
Access: API, Local
Parameters: 8.1B
Maximum output: 1 megapixels

Open model profile

Video models

23 models

Gemini Omni Flash

Released: June 2026

Weights: Proprietary
Access: API, Web
Maximum resolution: 720p
Maximum duration: 10 s

Open model profile

Black Forest Labs

FLUX.3 Video

Released: August 2026

Weights: Proprietary
Access: API, Web
Native audio: Yes

Open model profile

Dreamina Seedance 2.5 720p

Released: July 2026

Weights: Proprietary
Access: Web
Maximum resolution: 720p
Maximum duration: 30 s

Open model profile

Dreamina Seedance 2.0 720p

Released: March 2026

Weights: Proprietary
Access: API, Web
Maximum resolution: 720p
Maximum duration: 15 s

Open model profile

MiniMax H3

Released: July 2026

Weights: Open weights
Access: API, Web, Local
Parameters: 33B
Maximum resolution: 2K

Open model profile

Sora 2 Pro

Released: September 2025

Weights: Proprietary
Access: API
Maximum resolution: 1080p
Maximum duration: 20 s

Open model profile

Video modelLegacy

Sora 2

Released: September 2025

Weights: Proprietary
Access: API
Maximum resolution: 720p
Maximum duration: 20 s

Open model profile

Veo 3.1

Released: October 2025

Weights: Proprietary
Access: API, Web
Maximum resolution: 1080p
Maximum duration: 8 s

Open model profile

Veo 3.1 Fast

Released: October 2025

Weights: Proprietary
Access: API, Web
Maximum resolution: 1080p
Maximum duration: 8 s

Open model profile

Grok Imagine Video 1.5

Released: May 2026

Weights: Proprietary
Access: API, Web
Maximum resolution: 1080p
Native audio: Yes

Open model profile

Grok Imagine Video

Released: January 2026

Weights: Proprietary
Access: API, Web
Maximum resolution: 720p
Native audio: Yes

Open model profile

Wan 2.7

Released: April 2026

Weights: Proprietary
Access: API
Maximum resolution: 1080p
Maximum duration: 15 s

Open model profile

Wan 2.6

Released: December 2025

Weights: Proprietary
Access: API
Maximum resolution: 1080p
Maximum duration: 15 s

Open model profile

Runway Gen-4.5

Released: December 2025

Weights: Proprietary
Access: Web
Maximum resolution: 720p
Maximum duration: 10 s

Open model profile

Kling 3.0 Pro

Released: February 2026

Weights: Proprietary
Access: API, Web
Maximum resolution: 1080p
Maximum duration: 15 s

Open model profile

Kling 2.6 Pro

Released: January 2026

Weights: Proprietary
Access: API, Web
Maximum resolution: 1080p
Native audio: Yes

Open model profile

Vidu Q3 Pro

Released: January 2026

Weights: Proprietary
Access: API, Web
Maximum resolution: 1080p
Maximum duration: 16 s

Open model profile

PixVerse V5.6

Released: January 2026

Weights: Proprietary
Access: API, Web
Native audio: Yes

Open model profile

Hailuo 2.3

Released: October 2025

Weights: Proprietary
Access: API, Web
Maximum resolution: 1080p
Maximum duration: 10 s

Open model profile

LTX-2 19B

Released: October 2025

Weights: Open weights
Access: API, Web, Local
Parameters: 19B
Native audio: Yes

Open model profile

Wan 2.2 A14B

Released: July 2025

Weights: Open weights
Access: API, Local
Parameters: 27B
Active parameters: 14B

Open model profile

HunyuanVideo 1.5

Released: November 2025

Weights: Open weights
Access: API, Local
Parameters: 8.3B
Maximum resolution: 1080p (VSR)

Open model profile

Pika 2.2

Released: February 2025

Weights: Proprietary
Access: API, Web

Open model profile

Continue exploring

Comparisons and directories

Profiles describe individual entities. Use these pages for shared measurements or open-weight models.