announcement aidev SIG 4/5

Qwen open-sources Qwen2.5-1M models with 1M-token context windows

Qwen released two open-source checkpoints — Qwen2.5-7B-Instruct-1M and Qwen2.5-14B-Instruct-1M — upgraded to support context lengths of up to one million tokens. The release includes inference framework support. This follows the earlier Qwen2.5-Turbo upgrade to 1M context and marks the first time the open-source Qwen models have received this capability. Available on Hugging Face and ModelScope.

PUBLISHED2025-01-26
OBSERVED2026-08-11
AGE1y
SOURCES1
  • Two new open-source checkpoints: Qwen2.5-7B-Instruct-1M and Qwen2.5-14B-Instruct-1M
  • First open-source Qwen models with 1M-token context support
  • Inference framework support included in the release
  • Follows Qwen2.5-Turbo's earlier 1M context upgrade
  • Available on Hugging Face and ModelScope (per platform badges in source)
new-modelfeaturecontext-window

COMMUNITY

No curated reactions recorded for this event. Facts and takes are kept in separate layers — community context is added by hand, never blended into the record above.