LIVE
OpenAI models are coming to Google Cloud21/10/26 · OpenAI|OpenAI details ChatGPT's new advertising infrastructure21/10/26 · OpenAI|Anthropic reportedly posted a net loss of over $80 billion in H1 202620/10/26 · Anthropic|Self-Evolving Interfaces Help a Mid-Sized Open Model Beat GPT-5.5 and Claude Opus20/10/26|Sweden bans ChatGPT and Claude across government offices23/10/26|SpectralCanary: A Lightweight Method to Detect Unauthorized Training Data Use22/10/26|Low-bit n-gram tables cut memory use in language models22/10/26|NVIDIA releases OpenShell, a runtime for fleets of autonomous AI agents22/10/26 · NVIDIA|Ai2 releases OlmoEarth v1.3, an upgraded geospatial embeddings model22/10/26 · Ai2|NVIDIA Introduces Minerva, an Observability Platform for AI Factories22/10/26 · NVIDIA|Google DeepMind maps the world's forest types at global scale22/10/26 · Google DeepMind|AI's hidden storage bill: what inference leaves behind22/10/26|OpenAI models are coming to Google Cloud21/10/26 · OpenAI|OpenAI details ChatGPT's new advertising infrastructure21/10/26 · OpenAI|Anthropic reportedly posted a net loss of over $80 billion in H1 202620/10/26 · Anthropic|Self-Evolving Interfaces Help a Mid-Sized Open Model Beat GPT-5.5 and Claude Opus20/10/26|Sweden bans ChatGPT and Claude across government offices23/10/26|SpectralCanary: A Lightweight Method to Detect Unauthorized Training Data Use22/10/26|Low-bit n-gram tables cut memory use in language models22/10/26|NVIDIA releases OpenShell, a runtime for fleets of autonomous AI agents22/10/26 · NVIDIA|Ai2 releases OlmoEarth v1.3, an upgraded geospatial embeddings model22/10/26 · Ai2|NVIDIA Introduces Minerva, an Observability Platform for AI Factories22/10/26 · NVIDIA|Google DeepMind maps the world's forest types at global scale22/10/26 · Google DeepMind|AI's hidden storage bill: what inference leaves behind22/10/26|

Which AI model for your need?

Three to six questions. Each answer narrows the list, live, on the comparator’s verified data. No model is picked by an AI: the ranking is deterministic and every reason is a fact.

Typical paths

Nine common needs, already answered

Same rules as the Finder, computed on today’s data. Click a path to resume and adjust it.

Code locally on a 24 GB card

For meCodeOn my machinePC with 24 GB card
20 models pass these constraints
  1. 1
    Arena Elo 1438 · +40 use-case fit · #1 within your constraints · Runs in Q4 on your machine (19 GB)
    Context 262k · apache-2.0
  2. 2
    Arena Elo 1452 · +15 use-case fit · #2 within your constraints · Runs in Q4 on your machine (21 GB)
    Context 256k · apache-2.0
  3. 3
    Devstral SmallMistral AI· Different vendor
    ≈ Elo 1321 estimated from Artificial Analysis · +40 use-case fit · #6 within your constraints · Runs in Q4 on your machine (17 GB)
    Context 128k · apache-2.0
Resume this path

An assistant on a laptop without GPU

For meChat, think, searchOn my machineLaptop 8 GB, no dedicated GPU
16 models pass these constraints
  1. 1
    ≈ Elo 1373 estimated from Artificial Analysis · +40 use-case fit · #1 within your constraints · Runs in Q4 on your machine (4 GB)
    Context 131k · Apache-2.0
  2. 2
    ≈ Elo 1321 estimated from Artificial Analysis · #2 within your constraints · Runs in Q4 on your machine (5 GB)
    Context 128k · apache-2.0
  3. 3
    Qwen3 4BAlibaba· Different vendor
    ≈ Elo 1288 estimated from Artificial Analysis · #5 within your constraints · Runs in Q4 on your machine (5 GB)
    Context 32k · apache-2.0
Resume this path

Code via API, normal budget

For meCodeOnline, via APINormal
22 models pass these constraints
  1. 1
    Arena Elo 1466 · +40 use-case fit · #1 within your constraints · Code-specialised
    Context 1.1M · Output / M $12.00 · proprietary
  2. 2
    Arena Elo 1488 · +15 use-case fit · #2 within your constraints · Code-specialised
    Context 1.0M · Output / M $3.75 · proprietary
  3. 3
    Kimi K2.6Moonshot AI
    Arena Elo 1461 · +40 use-case fit · #3 within your constraints · Code-specialised
    Context 262k · Output / M $4.00 · modified-mit
Resume this path

Generate images online

For meGenerate imagesOnline, via APINo constraint
12 models pass these constraints
  1. 1
    Arena Elo text-to-image 1425 · #1 within your constraints
    · proprietary
  2. 2
    MAI-Image 2.6Microsoft AI· Different vendor
    Arena Elo text-to-image 1332 · #3 within your constraints
    · proprietary
  3. 3
    Reve 2.1Reve· Different vendor
    Arena Elo text-to-image 1302 · #4 within your constraints
    · proprietary
Resume this path

RAG on internal documents, EU hosting

For my team or companyAnalyse our documentsInternalWith an EU hostOne departmentNormal
110 models pass these constraints
  1. 1
    Arena Elo 1496 · +20 use-case fit · #1 within your constraints · EU hosting option from the vendor
    Context 1.0M · Output / M $3.75 · proprietary
  2. 2
    Mistral Medium 3.5Mistral AI· Best European
    Arena Elo 1426 · +20 use-case fit · #37 within your constraints · European vendor (Mistral AI)
    Context 262k · Output / M $7.50 · modified-mit
  3. 3
    Kimi K3Moonshot AI· Different vendor
    Arena Elo 1488 · +20 use-case fit · #2 within your constraints · 1.0M context, comfortable for RAG
    Context 1.0M · kimi-k3
Resume this path

RAG on sensitive data, on your servers

For my team or companyAnalyse our documentsSensitive or regulatedOn our serversGPU serverThe whole company
77 models pass these constraints
  1. 1
    Kimi K3Moonshot AI
    Arena Elo 1488 · +20 use-case fit · #1 within your constraints · Runs in FP16 on a GPU server (est. 6162 GB VRAM)
    Context 1.0M · kimi-k3
  2. 2
    Mistral Medium 3.5Mistral AI· Best European
    Arena Elo 1426 · +20 use-case fit · #20 within your constraints · Runs in FP16 on a GPU server (est. 284 GB VRAM)
    Context 262k · Output / M $7.50 · modified-mit
  3. 3
    MiMo-V2.6 ProXiaomi· Different vendor
    Arena Elo 1479 · +20 use-case fit · #2 within your constraints · Self-hostable: no data leaves
    Context 1.0M · mit
Resume this path

Agents and automation via API

For my team or companyAgents and automationPublic or anonymisedOnline, via APIOne departmentNormal
17 models pass these constraints
  1. 1
    Arena Elo 1494 · +40 use-case fit · #1 within your constraints · Built for tool calling and agents
    Context 1.0M · Output / M $4.25 · proprietary
  2. 2
    Mistral Large 4Mistral AI· Best European
    Arena Elo 1432 · +15 use-case fit · #13 within your constraints · Built for tool calling and agents
    · Output / M $4.18 · unknown
  3. 3
    ≈ Elo 1479 estimated from Artificial Analysis · +40 use-case fit · #3 within your constraints · Built for tool calling and agents
    Context 1.1M · Output / M $10.00 · proprietary
Resume this path

Translation and multilingual, tight budget

For my team or companyTranslation and multilingualInternalCloud is fineOne departmentTight
45 models pass these constraints
  1. 1
    MiMo-V2.6 ProXiaomi· Top ranked
    Arena Elo 1479 · #1 within your constraints
    Context 1.0M · mit
  2. 2
    Qwen3.7 PlusAlibaba· Best via API
    Arena Elo 1456 · #5 within your constraints · API hosted in China
    Context 1M · Output / M $1.60 · proprietary
  3. 3
    GPT-5.6 LunaOpenAI· Best value
    Arena Elo 1453 · #7 within your constraints
    Context 1.1M · Output / M $1.20 · proprietary
Resume this path

High-volume customer service

For my team or companyCustomer servicePublic or anonymisedOnline, via APIThe whole companyTight
5 models pass these constraints
  1. 1
    Arena Elo 1456 · #1 within your constraints · API hosted in China
    Context 1M · Output / M $1.60 · proprietary
  2. 2
    Mistral Large 3Mistral AI· Best European
    Arena Elo 1414 · +15 use-case fit · #3 within your constraints · First-class French
    Context 256k · Output / M $1.50 · apache-2.0
  3. 3
    GPT-5.6 LunaOpenAI· Different vendor
    Arena Elo 1453 · #2 within your constraints
    Context 1.1M · Output / M $1.20 · proprietary
Resume this path