📌OpenAI GPT-5.5 & DeepSeek V4 launched simultaneously


🇺🇸OpenAI releases GPT-5.5 (23/4/2026)
- GPT-5.5 codenamed "Spud," called by Greg Brockman "a new layer of intelligence aimed at more agentic and intuitive computing."
- GPT 5.5 is strongest in agentic tasks: programming, computer use, office work, and scientific research.
- API price: $5 per million input tokens and $30 per million output tokens. The Pro version is much more expensive: $30 and $180.
- GPT-5.5 Thinking leads most agentic tests (Terminal-Bench, OSWorld, CyberGym), while GPT-5.5 Pro excels in BrowseComp (90.1%) and FrontierMath.
-> Launch just 6 weeks after GPT-5.4, reflecting rapid release pace driven by fierce competition for enterprise customers.
🇨🇳 DeepSeek launches V4 Pro & Flash (24/4/2026)
- DeepSeek introduces two V4 Mixture-of-Experts models with a 1 million token context window. V4-Pro has 1.6 trillion parameters (49B active) — becoming the largest open-weight model to date. V4-Flash is smaller with 284B parameters (13B active).
- V4-Pro tops all open-source models in math and programming, only behind Gemini 3.1-Pro (closed-source) in world knowledge.
- DeepSeek aggressively prices V4-Pro at only $3.48 per million output tokens — compared to OpenAI $30 and Anthropic $25 for the same volume.
-> Performance "close to" GPT-5.4 but claims to be "about 3–6 months behind frontier." For the first time, DeepSeek trains models using Huawei Ascend chips instead of Nvidia. China begins to have alternatives to avoid US chips in the future.
View Original
post-image
This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.
  • Reward
  • Comment
  • Repost
  • Share
Comment
Add a comment
Add a comment
No comments
  • Pin