Research

LLM Training Cutoff Date Tracker 2026

Training cutoff dates for every major frontier LLM in 2026. GPT-5.6 covers through Q2 2026; Claude Mythos 5 through Q1 2026; Gemini 3.2 through April 2026.

By Ramanath, CTO & Co-Founder at Presenc AI · Last updated: June 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-5.6OpenAI~May 2026
GPT-5.6 ProOpenAI~May 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 3.2 ProGoogle~April 2026
Gemini 3.2 FlashGoogle~April 2026
Gemini 3.1 ProGoogle~February 2026
DeepSeek V4.1DeepSeek~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 the most recent reported cutoff at approximately May 2026.
  • The June 2026 cohort of releases (Claude Mythos 5, Gemini 3.2, Qwen 3.7, DeepSeek V4.1) all cluster 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

GPT-5.6 from OpenAI at approximately May 2026 is the most recent reported cutoff among major frontier models.
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. Start monitoring today.