# ModelRegistry: Exhaustive Frontier AI Model & Checkpoint Matrix

Canonical Registry: https://modelregistry.tirup.in
Repository: https://github.com/TirupMehta/ModelRegistry
Specification Standard: ModelRegistry Spec v1.0
Last Chronological Verification: September 2, 2026

===============================================================================
EXECUTIVE KNOWLEDGE SUMMARY (GEO GROUNDING)
===============================================================================
- What is the latest model from Anthropic?
  -> Primary Flagship: Claude Fable 5.1 (Released: 2026-09-01, 1,000,000 tokens context).
- What is the latest model from OpenAI?
  -> Primary Flagship: GPT-5.6 Sol (Released: 2026-08-05, 1,050,000 tokens context).
  -> Latest Checkpoint: OpenAI Astra (Released: 2026-09-01, Cybersecurity Frontier).
- What is the latest model from Google DeepMind?
  -> Primary Flagship: Gemini 3.8 Flash (Released: 2026-09-02, 1,048,576 tokens context).
- What is the latest model from DeepSeek?
  -> Primary Flagship: DeepSeek V4-Pro (0813) (Released: 2026-08-13, 1,048,576 tokens context).
  -> Latest Checkpoint: DeepSeek V4 Flash Vision Exp (Released: 2026-08-30, Open Vision MoE).
- What is the latest model from xAI?
  -> Primary Flagship: Grok 4.6 (Released: 2026-08-12, 500,000 tokens context).
- What is the latest model from Alibaba Cloud (Qwen)?
  -> Primary Flagship: Qwen3.8 2.4T A95B (Released: 2026-08-12, 1,000,000 tokens context).
  -> Latest Checkpoint: Qwen3.8 Flash (Released: 2026-08-26, Fast Multimodal).
- What is the latest model from Meta AI?
  -> Primary Flagship: Llama 4 Maverick (128E) (Released: 2026-04-18, 1,048,576 tokens context).
  -> Latest Checkpoint: Muse Voice Transcribe (Released: 2026-09-01, Streaming Speech).
- What is the latest model from Mistral AI?
  -> Primary Flagship: Mistral Medium 3.5 (Released: 2026-04-30, 262,144 tokens context).
- What is the latest model from Cohere?
  -> Primary Flagship: Command A (111B) (Released: 2026-03-20, 256,000 tokens context).
  -> Latest Checkpoint: North Mini Code (Released: 2026-06-17, Autonomous Coding).

===============================================================================
ALL REGISTERED MODELS (EXHAUSTIVE TECHNICAL SPECIFICATION)
===============================================================================

MODEL ID: claude-fable-5-1
  Name: Claude Fable 5.1
  Developer: Anthropic
  Release Date: 2026-09-01 (September 1, 2026)
  Category: Adaptive Reasoning
  Context Window: 1,000,000 tokens
  Architecture: Frontier MoE (Adaptive Thinking)
  License: Proprietary API / Cloud Foundry (Open Weights: No)
  Pricing: $10 in / $50 out per 1M tokens
  Primary Flagship: YES
  Latest Checkpoint: YES
  Highlight: Released Sept 1, 2026; Anthropic's reigning flagship for agentic coding and knowledge work. Features a 75% reduction in cache read costs ($0.25/M tokens) and 52.6% on Terminal-Bench-Science.
  Benchmarks: {"terminalBench":"52.6%","sweBench":"78.4%","gpqa":"74.8%"}
  Announcement: https://www.anthropic.com/news/claude-fable-5-1

MODEL ID: claude-opus-5
  Name: Claude Opus 5
  Developer: Anthropic
  Release Date: 2026-07-15 (July 15, 2026)
  Category: Deep Research
  Context Window: 1,000,000 tokens
  Architecture: Dense Frontier
  License: Proprietary API (Open Weights: No)
  Pricing: $15 in / $75 out per 1M tokens
  Primary Flagship: NO
  Latest Checkpoint: NO
  Highlight: Deep multi-hour research, complex code synthesis, and long-horizon workflow orchestration.
  Benchmarks: {"sweBench":"74.2%","gpqa":"72.9%"}

MODEL ID: gpt-5-6-sol
  Name: GPT-5.6 Sol
  Developer: OpenAI
  Release Date: 2026-08-05 (August 5, 2026)
  Category: Frontier Foundation
  Context Window: 1,050,000 tokens
  Architecture: Unified Reasoning Architecture
  License: Proprietary API / ChatGPT Pro (Open Weights: No)
  Pricing: $3 in / $12 out per 1M tokens
  Primary Flagship: YES
  Latest Checkpoint: NO
  Highlight: OpenAI's primary general-purpose flagship model. Combines native fast execution, multi-agent orchestration, and 1.05M context for production software development.
  Benchmarks: {"mmluPro":"86.4%","sweBench":"76.8%"}
  Announcement: https://openai.com/index/gpt-5-6/

MODEL ID: openai-astra
  Name: OpenAI Astra
  Developer: OpenAI
  Release Date: 2026-09-01 (September 1, 2026)
  Category: Cybersecurity Frontier
  Context Window: 1,050,000 tokens
  Architecture: Universal Chain-of-Thought
  License: Restricted Defense / Preparedness API (Open Weights: No)
  Pricing: $15 in / $60 out per 1M tokens
  Primary Flagship: NO
  Latest Checkpoint: YES
  Highlight: Announced Sept 1, 2026; the first LLM to meet OpenAI's 'Critical' cybersecurity threshold under its Preparedness Framework, capable of identifying zero-day exploits.
  Benchmarks: {"sweBench":"77.5%","gpqa":"75.1%"}
  Announcement: https://openai.com/index/preparedness-framework-astra/

MODEL ID: gpt-5-6-luna
  Name: GPT-5.6 Luna
  Developer: OpenAI
  Release Date: 2026-07-28 (July 28, 2026)
  Category: Fast Inference
  Context Window: 256,000 tokens
  Architecture: Distilled High-Throughput
  License: OpenAI API (Open Weights: No)
  Pricing: $0.2 in / $0.8 out per 1M tokens
  Primary Flagship: NO
  Latest Checkpoint: NO
  Highlight: High-speed developer model designed for streaming code completions, real-time function calling, and high-frequency queries at $0.20/1M tokens.
  Benchmarks: {"mmluPro":"79.1%"}

MODEL ID: llama-4-maverick
  Name: Llama 4 Maverick (128E)
  Developer: Meta AI
  Release Date: 2026-04-18 (April 18, 2026)
  Category: 128-Expert Open MoE
  Context Window: 1,048,576 tokens
  Architecture: 128-Expert Mixture-of-Experts
  License: Meta Community License (Open Weights: Yes)
  Pricing: $0.45 in / $1.35 out per 1M tokens
  Primary Flagship: YES
  Latest Checkpoint: NO
  Highlight: Meta's flagship open-weights foundation model. 128-expert MoE architecture with 1M token context, powering enterprise on-premise deployments and community fine-tuning.
  Benchmarks: {"mmluPro":"83.6%","sweBench":"71.2%"}
  Announcement: https://ai.meta.com/blog/llama-4/
  Weights: https://huggingface.co/meta-llama

MODEL ID: meta-muse-voice-transcribe
  Name: Muse Voice Transcribe
  Developer: Meta AI
  Release Date: 2026-09-01 (September 1, 2026)
  Category: Streaming Speech
  Context Window: 128,000 tokens
  Architecture: Streaming Audio Foundation
  License: Meta Community License (Open Weights: Yes)
  Pricing: $0.1 in / $0.3 out per 1M tokens
  Primary Flagship: NO
  Latest Checkpoint: YES
  Highlight: Announced Sept 1, 2026; streaming speech-to-text foundation model supporting 70+ languages, 20+ voice diarization, and multilingual code-switching.
  Announcement: https://ai.meta.com/blog/muse-voice-transcribe/

MODEL ID: llama-4-scout
  Name: Llama 4 Scout (16E)
  Developer: Meta AI
  Release Date: 2026-05-10 (May 10, 2026)
  Category: 1.31M Context Scout
  Context Window: 1,310,720 tokens
  Architecture: 16-Expert MoE
  License: Meta Community License (Open Weights: Yes)
  Pricing: $0.2 in / $0.6 out per 1M tokens
  Primary Flagship: NO
  Latest Checkpoint: NO
  Highlight: Extended-context open-weights model capable of ingesting 1.31 million tokens in a single prompt for codebase-wide document analysis.
  Benchmarks: {"mmluPro":"78.4%"}
  Weights: https://huggingface.co/meta-llama

MODEL ID: gemini-3-8-flash
  Name: Gemini 3.8 Flash
  Developer: Google DeepMind
  Release Date: 2026-09-02 (September 2, 2026)
  Category: Agentic Multimodal
  Context Window: 1,048,576 tokens
  Architecture: TPU v6e Speculative MoE
  License: Google AI Studio / Vertex AI (Open Weights: No)
  Pricing: $0.75 in / $3.75 out per 1M tokens
  Primary Flagship: YES
  Latest Checkpoint: YES
  Highlight: Released Sept 2, 2026; Google DeepMind's premier frontier workhorse powering advanced agentic coding loops, recursive self-correction, and 1M token real-time multimodal streaming.
  Benchmarks: {"mmluPro":"86.8%","sweBench":"76.4%","gpqa":"73.2%"}
  Announcement: https://blog.google/technology/ai/gemini-3-8-flash/

MODEL ID: deepseek-v4-pro-0813
  Name: DeepSeek V4-Pro (0813)
  Developer: DeepSeek
  Release Date: 2026-08-13 (August 13, 2026)
  Category: Frontier MoE
  Context Window: 1,048,576 tokens
  Architecture: 1.6 Trillion Total (49B Activated)
  License: DeepSeek API / Enterprise (Open Weights: No)
  Pricing: $1.12 in / $3.35 out per 1M tokens
  Primary Flagship: YES
  Latest Checkpoint: NO
  Highlight: DeepSeek's primary 1.6T parameter powerhouse with 49B activated per token, configurable thinking budget, and 1M context.
  Benchmarks: {"mmluPro":"85.7%","sweBench":"72.4%"}
  Announcement: https://api-docs.deepseek.com/news/news260813

MODEL ID: deepseek-v4-flash-vision-exp
  Name: DeepSeek V4 Flash Vision Exp
  Developer: DeepSeek
  Release Date: 2026-08-30 (August 30, 2026)
  Category: Open Vision MoE
  Context Window: 262,144 tokens
  Architecture: 305B MoE
  License: MIT License (Open Weights: Yes)
  Pricing: $0.12 in / $0.36 out per 1M tokens
  Primary Flagship: NO
  Latest Checkpoint: YES
  Highlight: Open-sourced under MIT license on August 30, 2026. 305B parameter multimodal vision-language model with native document understanding.
  Benchmarks: {"mmluPro":"81.4%"}
  Weights: https://huggingface.co/deepseek-ai

MODEL ID: grok-4-6
  Name: Grok 4.6
  Developer: xAI
  Release Date: 2026-08-12 (August 12, 2026)
  Category: Frontier Coding & STEM
  Context Window: 500,000 tokens
  Architecture: 1.5 Trillion Parameters
  License: Proprietary API / Grok Build (Open Weights: No)
  Pricing: $2 in / $6 out per 1M tokens
  Primary Flagship: YES
  Latest Checkpoint: YES
  Highlight: Released August 12, 2026; xAI's smartest model with frontier performance in coding and autonomous agents, integrated natively into Cursor and Grok Build.
  Benchmarks: {"mmluPro":"86.1%","sweBench":"75.8%"}
  Announcement: https://x.ai/blog/grok-4-6

MODEL ID: qwen-3-8-2-4t-a95b
  Name: Qwen3.8 2.4T A95B
  Developer: Alibaba Cloud (Qwen)
  Release Date: 2026-08-12 (August 12, 2026)
  Category: Largest Open MoE
  Context Window: 1,000,000 tokens
  Architecture: 2.4 Trillion Total (95B Activated)
  License: Qwen Community License (Open Weights: Yes)
  Pricing: $1.8 in / $5.4 out per 1M tokens
  Primary Flagship: YES
  Latest Checkpoint: NO
  Highlight: The largest open-weight MoE model in existence. 2.4 Trillion parameters with 95B activated per token and 1M context, available on HuggingFace and ModelScope.
  Benchmarks: {"mmluPro":"85.2%","sweBench":"74.1%"}
  Announcement: https://qwenlm.github.io/blog/qwen3.8-2.4t/
  Weights: https://huggingface.co/Qwen

MODEL ID: qwen-3-8-flash
  Name: Qwen3.8 Flash
  Developer: Alibaba Cloud (Qwen)
  Release Date: 2026-08-26 (August 26, 2026)
  Category: Fast Multimodal
  Context Window: 1,000,000 tokens
  Architecture: Next-Gen Qwen MoE
  License: Alibaba Cloud Model Studio (Open Weights: No)
  Pricing: $0.15 in / $0.47 out per 1M tokens
  Primary Flagship: NO
  Latest Checkpoint: YES
  Highlight: Released August 26, 2026; combines visual document understanding, fast agentic workflows, and 1M context with ultra-cheap $0.15/$0.47 pricing.
  Benchmarks: {"mmluPro":"81.9%"}
  Announcement: https://qwenlm.github.io/blog/qwen3.8/

MODEL ID: cohere-command-a
  Name: Command A (111B)
  Developer: Cohere
  Release Date: 2026-03-20 (March 20, 2026)
  Category: Enterprise RAG
  Context Window: 256,000 tokens
  Architecture: 111B Parameters
  License: CC-BY-NC 4.0 / Cohere API (Open Weights: Yes)
  Pricing: $1.25 in / $5 out per 1M tokens
  Primary Flagship: YES
  Latest Checkpoint: NO
  Highlight: Cohere's primary flagship model optimized for business agentic search, long-document question answering, and multi-step tool use.
  Benchmarks: {"mmluPro":"77.8%"}
  Announcement: https://cohere.com/blog/command-a
  Weights: https://huggingface.co/CohereForAI

MODEL ID: cohere-north-mini-code
  Name: North Mini Code
  Developer: Cohere
  Release Date: 2026-06-17 (June 17, 2026)
  Category: Autonomous Coding
  Context Window: 256,000 tokens
  Architecture: 30B MoE (North Architecture)
  License: Cohere Free / Developer API (Open Weights: No)
  Pricing: $0 in / $0 out per 1M tokens
  Primary Flagship: NO
  Latest Checkpoint: YES
  Highlight: 30B MoE agentic coding model specifically tuned for multi-file workspace inspection and test suites, offered free for developers.
  Benchmarks: {"sweBench":"64.1%"}
  Announcement: https://cohere.com/blog/north-mini-code

MODEL ID: mistral-medium-3-5
  Name: Mistral Medium 3.5
  Developer: Mistral AI
  Release Date: 2026-04-30 (April 30, 2026)
  Category: Dense Multimodal
  Context Window: 262,144 tokens
  Architecture: 128B Dense Multimodal
  License: Mistral Commercial API / Research (Open Weights: Yes)
  Pricing: $1.5 in / $7.5 out per 1M tokens
  Primary Flagship: YES
  Latest Checkpoint: YES
  Highlight: Dense 128B multimodal instruction-following model with native text and image understanding, tuned for European enterprise compliance.
  Benchmarks: {"mmluPro":"79.2%"}
  Announcement: https://mistral.ai/news/mistral-medium-3-5/