Skip to main content
idapt
AccueilModèles IATarifs
Sign inStart free
All roundups

Best of

Open weights at the frontier.

The open-weight models that stand next to closed flagships on the live rankings, hosted or on your own hardware.

Last reviewed July 16, 2026

Try idapt freeBrowse all models

The gap between open and closed models keeps narrowing: today's open-weight flagships sit on the same leaderboards as closed ones, and several picks below lead their lanes outright. Open weights buy you provider choice, local deployment, and price competition; here is where that trade is strongest.

How we pickedEvery pick publishes downloadable weights. Standing comes from the same live rankings the closed models compete on; prices render from the catalog's hosted rates, and the local-deployment notes point at the sizes that fit real hardware.

The picks

1DeepSeek V4 ProThe open flagship.

$0.43 in / $0.87 out per M tokens · 1.0M context

Frontier-lane benchmark standing with published weights; the strongest argument that open has caught up.

Best for: Frontier-grade work with open-weight requirements.

2Qwen 3.7 MaxThe broadest open family.

$1.25 in / $3.75 out per M tokens · 1M context

One family covering flagship, coder, reasoner, and small sizes, so one vocabulary of behavior scales from a laptop to a cluster.

Best for: Teams standardizing on one open family.

3Nemotron 3 UltraNVIDIA's open frontier reasoner.

$0.5 in / $2.20 out per M tokens · 262K context

A very large open MoE tuned for reasoning, competitive on the hard lanes and fully open.

Best for: Hard reasoning with open weights.

4GLM-5.2The bilingual open expert.

$1.40 in / $4.40 out per M tokens · 262K context

Strong across reasoning and coding lanes with first-class Chinese-English performance, openly released.

Best for: Bilingual products and mixed workloads.

5Gemma 4 31BThe local-size standard.

$0.12 in / $0.35 out per M tokens · 262K context

Google's open line at a size a workstation GPU can serve; the reference point for capable local inference.

Best for: Local-first setups and private data.

Frequently asked

Run them in one workspace

Local AI

Run open models on your own hardware with Ollama: private, free, no rate limits.

Multi-Model Chat

GPT, Claude, Gemini, Grok, Llama: pick mid-conversation, cheapest-first.

Keep comparing

Serious models, your hardware.Frontier answers at commodity prices.The models that ship working code.Real local AI on 8 GB.idapt vs Open WebUI

Judge them on your own prompt

Every pick on this page runs in one idapt chat. Start free with a daily allowance, no credit card.

Try idapt free
  • Accueil
  • Tarifs
  • Modèles IA
  • Modèles d'image
  • Modèles de voix
  • Modèles vidéo
  • Classements
  • New models
  • Model status
  • Multi-Model Chat
  • Voice Mode
  • Voice HUD
  • Web Search
  • Image Generation
  • Video Generation
  • Audio Generation
  • Transcription
  • Drive
  • Secrets
  • Sharing
  • Workspaces
  • Tasks
  • Memory
  • Agents
  • Subagents
  • Automations
  • Skills
  • Code Execution
  • Computers
  • Computer Use
  • Cloud Computers
  • Local AI
  • AI Gateway
  • API & SDK
  • CLI
  • MCP
  • Tunnels
  • All features →
  • Blog
  • Cas d'usage
  • Comparisons
  • Best of
  • Learn
  • Free tools
  • Prompts
  • Templates
  • Changelog
  • Centre d'aide
  • FAQ
  • Confidentialité
  • Comparer tous les modèles
  • Support
  • Développeurs
  • Quickstarts
  • Référence API
  • Tarifs API
  • CLI
  • MCP
  • Downloads
  • Desktop
  • Badges and embeds
© idapt[email protected]ConditionsPolitique de confidentialitéMentions légalesSignaler un contenu
X (Twitter)