Every model we test

AI Models

Who makes each model, what it is, and what it is best at, with scores from real reviewed results. Open a model for its full profile, or see the ranked scores on the Leaderboard.

TEXT MODELS28

Text modelsChat, writing, reasoning, and coding models

  • Qwen 3.8 2.4T

    Alibaba

    Alibaba's Qwen 3.8 2.4T-parameter mixture-of-experts model (95B active).

    • Not enough reviews to score yet

  • Qwen 3.8 Max

    Alibaba

    The top Qwen model for reasoning, coding, and multilingual work.

    • Not enough reviews to score yet

  • Claude Haiku 4.5

    Anthropic
    8.49/10

    Optimised for speed and efficiency.

    • Stands out in Accuracy
    • 18.9s average
    • 12 reviews

  • Claude Opus 4.6

    Anthropic
    8.41/10

    Advanced reasoning, coding, and research.

    • Stands out in Coherence
    • 46.4s average
    • 8 reviews

  • Claude Opus 5

    Anthropic

    The flagship Claude model for demanding reasoning, coding, and agent work.

    • Not enough reviews to score yet

  • Claude Sonnet 5

    Anthropic

    Strong agentic/coding performance at a lower cost than Opus.

    • Not enough reviews to score yet

  • DeepSeek V4 Pro

    DeepSeek

    1.6T-parameter MoE reasoning model, successor to R1.

    • Not enough reviews to score yet

  • DeepSeek V4.1 Flash

    DeepSeek

    The fast, efficient member of the V4.1 family.

    • Not enough reviews to score yet

  • Gemini 3.1 Pro

    Google

    1M-token context, advanced reasoning across text, audio, image, video, and code.

    • Not enough reviews to score yet

  • Gemma 4 26B MoE

    Google
    8.78/10

    4B active parameters, fast and efficient.

    • Stands out in Creativity
    • 41.4s average
    • 14 reviews

  • Gemma 4 31B

    Google
    8.72/10

    256K context, multimodal, 140+ languages, agentic capabilities.

    • Stands out in Creativity
    • 51.6s average
    • 12 reviews

  • Llama 4 Maverick

    Meta

    17B active params, 128-expert MoE, multimodal, strong reasoning.

    • Not enough reviews to score yet

  • Llama 4 Scout

    Meta
    8.0/10

    Meta's Llama 4 Scout 17B instruction-tuned model.

    • Stands out in Relevance
    • 13.0s average
    • 14 reviews

  • Muse Glimmer 30B

    Meta

    A compact Muse model from Meta Superintelligence Labs.

    • Not enough reviews to score yet

  • MAI Thinking 1

    Microsoft

    Microsoft's first-party MAI Thinking 1 reasoning model.

    • Not enough reviews to score yet

  • Phi-4

    Microsoft
    7.76/10

    Efficient and capable.

    • Stands out in Language Quality
    • 30.0s average
    • 9 reviews

  • Mistral Medium

    Mistral AI
    7.94/10

    Mistral AI's balanced medium-sized model (2505 release).

    • Stands out in Coherence
    • 36.6s average
    • 8 reviews

  • Kimi K2.6

    Moonshot AI

    1T-parameter MoE, long-horizon agentic coding and tool use.

    • Not enough reviews to score yet

  • Kimi K3

    Moonshot AI

    The successor to K2.6 for long-horizon agentic coding and tool use.

    • Not enough reviews to score yet

  • Nemotron 3 Ultra

    NVIDIA

    The largest Nemotron 3 model for reasoning and agentic tasks.

    • Not enough reviews to score yet

  • Nemotron Lightning 3.5

    NVIDIA

    A fast, efficient Nemotron.

    • Not enough reviews to score yet

  • GPT-4.1 Mini

    OpenAI
    8.26/10

    OpenAI's fast and cost-efficient GPT-4.1 Mini.

    • Stands out in Accuracy
    • 9.5s average
    • 10 reviews

  • GPT-5.4

    OpenAI
    8.25/10

    Advanced reasoning, coding, and professional workflows.

    • Stands out in Coherence
    • 9.8s average
    • 9 reviews

  • GPT-5.6

    OpenAI

    Advanced reasoning, coding, and professional workflows.

    • Not enough reviews to score yet

  • Grok 4.1 Fast

    xAI
    7.81/10

    XAI's Grok 4.1 model optimised for fast agentic reasoning.

    • Stands out in Creativity
    • 15.9s average
    • 9 reviews

  • Grok 4.6

    xAI

    Long-horizon reasoning and complex workflows with a very large context window.

    • Not enough reviews to score yet

  • GPT OSS 120B

    Text model
    7.5/10

    Open-source 120B parameter GPT-class model.

    • Stands out in Creativity
    • 8.4s average
    • 9 reviews

  • DeepSeek R1

    DeepSeek · no longer offered
    8.41/10

    DeepSeek's flagship reasoning model with chain-of-thought capabilities.

    • Stands out in Creativity
    • 24.3s average
    • 12 reviews
IMAGE MODELS14

Image modelsImage generation models

  • FLUX.1 Dev

    Black Forest Labs
    8.65/10

    Fast, high-quality open image generation.

    • Stands out in Accuracy
    • 11.0s average
    • 4 reviews

  • FLUX.2 Pro

    Black Forest Labs

    Next-generation image quality and detail.

    • Not enough reviews to score yet

  • Gemini 3 Pro Image

    Google

    Reasoning-driven, studio-quality generation and editing. Premium successor to Imagen 4 Ultra.

    • Not enough reviews to score yet

  • Gemini 3.1 Flash Image

    Google
    8.54/10

    Native multimodal image generation, Google's recommended replacement for Imagen 4.

    • Stands out in Coherence
    • 10.1s average
    • 3 reviews

  • Gemini 3.1 Flash-Lite Image

    Google

    The fastest and cheapest native Gemini image generation.

    • Not enough reviews to score yet

  • MAI Image 2.5

    Microsoft

    Microsoft's MAI Image 2.5: photorealistic generation, the standard model of the MAI family.

    • Not enough reviews to score yet

  • MAI Image 2.5 Pro

    Microsoft

    The higher-fidelity MAI Image 2.5 variant.

    • Not enough reviews to score yet

  • MAI Image 2.6

    Microsoft

    Higher-quality photorealistic generation from the MAI family.

    • Not enough reviews to score yet

  • MAI Image 2.6 Flash

    Microsoft

    The fast variant of MAI Image 2.6.

    • Not enough reviews to score yet

  • GPT Image 1

    OpenAI

    State-of-the-art image generation with high prompt fidelity.

    • Not enough reviews to score yet

  • GPT Image 1 Mini

    OpenAI
    8.76/10

    Fast, high-quality image generation.

    • Stands out in Accuracy
    • 31.7s average
    • 5 reviews

  • GPT Image 2

    OpenAI

    Improved text rendering, face consistency, and instruction following.

    • Not enough reviews to score yet

  • GPT Image 2.5 Flare

    OpenAI

    The faster, lighter model for everyday image generation.

    • Not enough reviews to score yet

  • GPT Image 2.5 Sunburst

    OpenAI

    The heavier model, for demanding generations and precise edits.

    • Not enough reviews to score yet

Want the ranking? See the Leaderboard or run your own comparison.