Research

LLM Training Cutoff Date Tracker 2026

Training cutoff dates for every major frontier LLM in 2026. GPT-5.6 through February 16, 2026; GPT-6 Astra through April 30, 2026; Claude Opus 5 and Grok 4.7 through May 2026.

By Ramanath, CTO & Co-Founder at Presenc AI · Last updated: October 2026

This page tracks training cutoff dates for every major frontier LLM as of June 2026. Training cutoff determines which web events and brand mentions enter the parametric recall of the model.

Training Cutoff Snapshot (June 2026)

ModelVendorReported Cutoff
GPT-5.6OpenAIFebruary 16, 2026
GPT-5.6 ProOpenAIFebruary 16, 2026
GPT-6 AstraOpenAIApril 30, 2026
Claude Opus 5AnthropicMay 2026
Grok 4.7xAIMay 2026
Claude Mythos 5Anthropic~April 2026
Claude Opus 4.7Anthropic~January 2026
Claude Sonnet 4.6Anthropic~November 2025
Claude Haiku 4.5Anthropic~September 2025
Gemini 3.1 ProGoogle~February 2026
DeepSeek V4DeepSeek~April 2026
DeepSeek V4DeepSeek~February 2026
Qwen 3.7Alibaba~April 2026
Qwen 3.6Alibaba~February 2026
Llama 4.5Meta~March 2026
Llama 4 (original)Meta~December 2025
Mistral Medium 3Mistral AI~April 2026
Hunyuan Large 3Tencent~April 2026
ERNIE 5.1Baidu~April 2026
GLM-6Zhipu AI~April 2026
Grok 4xAI~Continuous via X integration

Key Takeaways

  • GPT-5.6 has a February 16, 2026 cutoff according to OpenAI. Later models moved the frontier forward: GPT-6 Astra to April 30, 2026, and Claude Opus 5 and Grok 4.7 to May 2026.
  • The mid-2026 cohort of releases (Claude Mythos 5, Qwen 3.7) clusters around an April 2026 cutoff.
  • Grok 4 uniquely incorporates continuous X-platform data through its native integration, with no fixed cutoff for X-derived content.
  • Brand mentions earned between two model versions' cutoffs enter the newer model's parametric recall and may exit older models that get deprecated.

Why Cutoff Dates Matter for Brand Visibility

Training cutoff determines what enters and exits a model's default-mode (parametric) recall. Brands that earned major coverage between the previous and current cutoff get added to recall. Retrieval-mode citation (ChatGPT Search, Perplexity, Google AI Overviews) is independent of cutoff and depends on which sources retrieval providers index.

Methodology

Cutoff dates from vendor disclosures, model cards, and reported benchmarking. Vendors disclose cutoff with varying precision; numbers expressed as approximate. Updated monthly.

How Presenc AI Helps

Presenc AI tracks brand visibility on each frontier model continuously so brand teams see when cutoff refreshes shift their default-mode recall.

Frequently Asked Questions

As of October 2026 the most recent stated cutoffs are May 2026 for Claude Opus 5 and Grok 4.7, and April 30, 2026 for GPT-6 Astra. GPT-5.6 has a February 16, 2026 cutoff.
Grok 4 uniquely incorporates continuous X-platform data through its native integration, with no fixed cutoff for X-derived content. Non-X training data follows a more conventional cutoff pattern.
No. Retrieval-mode citation (ChatGPT Search, Perplexity, Google AI Overviews) is independent of cutoff and depends on which sources retrieval providers index.
Approximately every six to eight weeks for the major Western frontier labs (OpenAI, Anthropic, Google) and approximately every two to three months for the Chinese frontier cohort.

Track Your AI Visibility

See how your brand appears across ChatGPT, Claude, Perplexity, and other AI platforms. Join the waitlist for early access.