Support

AI-powered help

Welcome!

Please introduce yourself before we start.

    Open Source Models

    Open-weight models — Llama, DeepSeek, Qwen, GLM, Kimi, GPT-OSS, Gemma, and more — served through one API

    Compare
    Use Case
    Capabilities
    Provider
    Status
    Input Price ($/M tokens)
    Output Price ($/M tokens)
    Context Size (tokens)
    92/323
    Models
    30/51
    Providers
    27
    Vision Models (filtered)
    70
    Tool-enabled (filtered)
    3
    Free Models (filtered)
    Features
    AWS Bedrock(us)
    llama-4-maverick-17b-instruct
    		$0.26$1.07—
    NovitaAI
    llama-4-maverick-17b-instruct
    		$0.27$0.85—
    SCX.ai (Turbo)
    llama-4-maverick-17b-instruct
    		$0.53$1.62—
    SCX.ai (Turbo)
    llama-4-maverick-17b-instruct
    		$0.53$1.62—
    AWS Bedrock
    llama-4-maverick-17b-instruct
    		$0.24$0.97—
    NovitaAI
    llama-4-scout-17b-instruct
    		$0.18$0.59—
    NovitaAI
    llama-4-scout-17b-instruct
    		$0.18$0.59—
    AWS Bedrock(us)
    llama-4-scout-17b-instruct
    		$0.19$0.73—
    AWS Bedrock
    llama-4-scout-17b-instruct
    		$0.17$0.66—
    AWS Bedrock
    llama-4-scout-17b-instruct
    		$0.17$0.66—
    Cerebras
    llama-3.3-70b-instruct
    		$0.85$1.20—
    Cerebras
    llama-3.3-70b-instruct
    		$0.85$1.20—
    Nebius AI
    llama-3.3-70b-instruct
    		$0.13$0.40—
    NovitaAI
    llama-3.3-70b-instruct
    		$0.14$0.40—
    NovitaAI
    llama-3.3-70b-instruct
    		$0.14$0.40—
    Nebius AI
    llama-3.3-70b-instruct
    		$0.13$0.40—
    Nebius AI
    llama-3.1-nemotron-ultra-253b
    		$0.60$1.80—
    Nebius AI
    llama-3.1-nemotron-ultra-253b
    		$0.60$1.80—
    Inference.net
    llama-3.2-11b-instruct
    		$0.07$0.33—
    Inference.net
    llama-3.2-11b-instruct
    		$0.07$0.33—
    AWS Bedrock(us)
    llama-3.1-70b-instruct
    		$0.79$0.79—
    AWS Bedrock
    llama-3.1-70b-instruct
    		$0.72$0.72—
    AWS Bedrock
    llama-3.1-70b-instruct
    		$0.72$0.72—
    Cerebras
    llama-3.1-8b-instruct
    		$0.10$0.10—
    Nebius AI
    llama-3.1-8b-instruct
    		$0.02$0.06—
    AWS Bedrock
    llama-3.1-8b-instruct
    		$0.22$0.22—
    NovitaAI
    llama-3.1-8b-instruct
    		$0.02$0.05—
    Xiaomi
    mimo-v2-flash
    		$0.10$0.30$0.02
    Xiaomi
    mimo-v2-omni
    		$0.40$2.00$0.08
    Xiaomi
    mimo-v2.5
    		$0.14$0.28$0.00
    Xiaomi
    mimo-v2.5
    		$0.14$0.28$0.00
    Xiaomi
    mimo-v2-pro
    		$1.00$3.00$0.20
    Xiaomi
    mimo-v2.5-pro
    		$0.43$0.87$0.00
    Xiaomi
    mimo-v2.5-pro
    		$0.43$0.87$0.00
    NovitaAI
    gemma-4-26b-a4b-it
    		$0.13$0.40—
    DeepInfra
    gemma-4-26b-a4b-it
    		$0.07$0.34—
    DeepInfra
    gemma-4-26b-a4b-it
    		$0.07$0.34—
    NovitaAI
    gemma-4-26b-a4b-it
    		$0.13$0.40—
    Cerebras
    gemma-4-31b-it
    		$0.99$1.49—
    Together AI
    gemma-4-31b-it
    		$0.39$0.97—
    DeepInfra
    gemma-4-31b-it
    		$0.13$0.38—
    Runware
    gemma-4-31b-it
    		$0.10$0.30$0.01
    NovitaAI
    gemma-4-31b-it
    		$0.14$0.40—
    Together AI
    gemma-4-31b-it
    		$0.39$0.97—
    Cerebras
    gemma-4-31b-it
    		$0.99$1.49—
    SCX.ai (Turbo)
    gemma-4-31b-it
    		$0.30$0.91—
    DeepInfra
    gemma-4-31b-it
    		$0.13$0.38—
    Runware
    gemma-4-31b-it
    		$0.10$0.30$0.01
    RanoAI
    gemma-4-31b-it
    		$0.10$0.30$0.05
    NovitaAI
    gemma-4-31b-it
    		$0.14$0.40—
    Page 8 of 9

    Newsletter

    Stay ahead of the curve

    Join developers who get weekly insights on LLM routing, new model launches, and cost optimization — straight to their inbox.

    • New models & providers as they drop
    • Tips to cut latency & costs
    • Early access to beta features

    No spam. Unsubscribe anytime.

    All systems operational
    AICPA SOC for Service Organizations badgeSOC 2 Type II
    compliant

    Product

    • Features
    • AI Gateway
    • Observability
    • Models
    • Providers
    • Rankings
    • Add Provider
    • Partners
    • Lounge
    • Changelog
    • Compare Models
    • Enterprise

    Resources

    • Legal Overview
    • Apps
    • MCP Server
    • Use Cases
    • Blog
    • Documentation
    • Integrations
    • Guides
    • Brand Assets
    • Token Cost Calculator
    • Copilot Cost Calculator
    • Referral Program
    • GitHub
    • Discord
    • Twitter
    • Contact Us

    Compliance

    • Trust Center
    • Security Portal
    • Terms
    • Privacy Policy
    • Provider Information
    • Sub-processors
    • SOC 2 Type II
    • Status

    Compare

    • All Comparisons
    • GitHub Copilot
    • OpenRouter
    • LiteLLM
    • Portkey
    • AWS Bedrock
    • Azure AI Foundry
    • Vercel AI Gateway
    • Migration Guides

    Models

    • Text Generation
    • Text to Image
    • Image to Image
    • Video Generation
    • Embeddings
    • Vision
    • Reasoning
    • Tool Calling
    • Web Search
    • Discounted
    • Best for Roleplay
    • Best for Coding
    • Best for Creative Writing
    • Best for Translation
    • Best for Math
    • Long Context
    • Cheapest
    • Open Source

    Providers

    • OpenAI
    • Anthropic
    • Google AI Studio
    • Google Vertex AI
    • Vertex AI (OpenAI-compatible)
    • Vertex AI (Anthropic)
    • Groq
    • Cerebras
    • xAI
    • DeepSeek
    • Alibaba Cloud
    • NovitaAI
    • AtlasCloud
    • AWS Bedrock
    • AWS Mantle
    • Azure
    • Azure AI Foundry
    • Z AI
    • Moonshot AI
    • Baidu
    • Perplexity
    • Nebius AI
    • Mistral AI
    • CanopyWave
    • Inference.net
    • Together AI
    • SCX.ai (Turbo)
    • SCX.ai
    • ByteDance
    • MiniMax
    • EmberCloud
    • Meta
    • Sakana AI
    • Xiaomi
    • DeepInfra
    • ElevenLabs
    • Runware
    • Gonka24
    • Fireworks AI

    © 2026 OffRail. All rights reserved.

    Open-weight models have closed most of the gap with proprietary frontiers: DeepSeek V4, Qwen3.7, GLM-5, Kimi K2, and MiniMax M3 sit near the top of real-world leaderboards, joined by OpenAI's GPT-OSS and Google's Gemma releases. Their weights are public — but running a 200B+ parameter model yourself means serious GPU infrastructure.

    This page lists open-weight models served by hosted providers, so you get the openness — inspectable weights, no lock-in, the option to self-host later — with API convenience.

    Frequently asked questions

    What is the best open source LLM?

    DeepSeek V4, Qwen3.7, GLM-5.2, Kimi K2.6, and MiniMax M3 are the current leaders, each within striking distance of proprietary frontier models. For smaller, hardware-friendly options, GPT-OSS 20B, Gemma 4, and Qwen3.5 9B are the standouts.

    What does 'open source' mean for LLMs?

    Usually 'open weight': the trained weights are downloadable, but licenses vary — some are Apache 2.0 or MIT, others (like the Llama license) carry usage restrictions, and training data is rarely published. Check the license of a specific model before building on it.

    Should I self-host or use an API?

    Self-hosting pays off with steady high volume, strict data-residency needs, or fine-tuned weights. For everything else, per-token APIs are cheaper than idle GPUs. A middle path: develop against hosted open models and keep self-hosting as an exit option, since the weights are public.

    Are open models cheaper than proprietary ones?

    Dramatically, per token. Competition among hosts drives prices down — DeepSeek V4 Flash and Qwen3 Coder 30B cost 10–50x less than frontier proprietary models. The list above shows every provider's price for each model.

    OffRail
    • Lounge
    • Models
    • Docs
    • Pricing
    • Lounge
    • Pricing
    • Docs
    • Models
      • AI Gateway
      • Lounge
      • Observability
      • Enterprise
      • Blog
      • Changelog
      • Integrations
      • Reliability
      • Guardrails
      • Providers
      • Partners
      • Rankings
      • Apps
      • Models
      • Model Timeline
      • Compare
      • Token Cost Calculator
      • Referral Program
      • MCP Server
      • AI SDK Provider
      • Guides
    Log InGet Started
    ★