New AI Model Releases May 2026 — Comprehensive Summary

Source Context: April 2026 was the most consequential month for frontier AI releases ever. In a single week (April 20–24), OpenAI shipped GPT-5.5, DeepSeek dropped V4 Preview, Moonshot AI launched Kimi K2.6, and xAI released Grok 4.3 — alongside Claude Opus 4.7 (April 16), Gemini 3.1 Pro, Llama 4, Qwen 3, and Gemma 4 within the same six-week window.

Model-by-Model Breakdown

1. Claude Mythos (Anthropic)

Status: Restricted preview (~50 partner organizations). No confirmed public release date. Training is complete. Capybara class model. Gated evaluations: 93.9% on SWE-bench Verified, 94.6% on GPQA Diamond. Project Glasswing: prioritizes cybersecurity organizations.

2. Meta Avocado

Status: Expected May or June 2026. Delayed from March 2026. Internal tests place it between Gemini 2.5 and Gemini 3.0. Planned proprietary model (departing from open-source Llama strategy). Meta planning $115–135B AI CAPEX in 2026.

3. DeepSeek V4 Full Release

Status: Preview live since April 24. Full open-weight release expected imminently. Architecture: V4 Pro 1.6T params (49B active), MoE; V4 Flash 284B params (13B active). Pricing: V4 Flash $0.14/M input, $0.28/M output; V4 Pro $3.48/M output. Undercuts GPT-5.5 ($30) and Claude Opus 4.7 ($25) by 7–9x. Text-only. DeepSeek seeking $20B valuation from Tencent/Alibaba.

4. Nemotron 4 (NVIDIA)

Status: Confirmed upcoming. Nemotron Coalition: Mistral AI, Black Forest Labs, Cursor, LangChain + 4 other labs. Nemotron 3 Nano Omni (April 29): 30B hybrid MoE, multimodal. Nemotron 4 follows after Super/Ultra release.

5. GR00T N2 (NVIDIA)

Status: Previewed at GTC March 2026. Expected by end of 2026. Vision Language Action (VLA) model for humanoid robotics.