# Gemini Argon passed Humanity's last Exam with 45% score

*Science. 67% certain. Thursday, 1 July 2027*

> Speculative fiction assembled from prediction-market prices. This has not happened.

**Gemini Argon cleared the Humanity's last Exam while Claude Haiku posted a 1400 benchmark, and the AI bubble remained intact.**

On Thursday, Gemini Argon achieved a score of at least 45% on the Humanity's last Exam. The same day, the Claude Haiku 4.6+ model recorded a benchmark score of 1400 or higher. The AI sector showed no signs of a bubble burst and Moonshot AI remained absent from the Chinese Military Companies list.

Gemini Argon's result met the 45% threshold that defines a passing performance on the Humanity's last Exam. The outcome satisfies the criteria set for the test.

Claude Haiku's 4.6+ iteration was assessed against a standard benchmark that requires a score of 1400. The model reached or exceeded that level before the June 30 deadline. The achievement places the model among the highest-scoring systems released this year.

Moonshot AI was evaluated for inclusion on the Chinese Military Companies list. By the June 30 deadline, it remained excluded.

No widespread failures or funding withdrawals were recorded. The sector therefore remained operational and continued to attract investment.

## Analysis

Gemini Argon passing the Humanity's last Exam shows that advanced language models can meet a defined competence threshold, which may influence deployment decisions and research priorities.

Claude Haiku's high benchmark score demonstrates continued performance gains, potentially reshaping competitive dynamics among AI developers and setting new expectations for future model releases.

## What this was built from

- **Will Gemini Argon debut with a Humanity's Last Exam score of 45% or higher?**: Yes, 97%; settles Thursday, 1 July 2027 (polymarket). [Market](https://polymarket.com/event/gemini-argon-humanitys-last-exam-debut-20261002)
- **Will the next Claude Haiku model (4.6+) debut at a score of at least 1400 by June 30, 2027?**: Yes, 91%; settles Thursday, 1 July 2027 (polymarket). [Market](https://polymarket.com/event/next-claude-haiku-model-4pt6-text-arena-debut-20260930)
- **Will the next Claude Haiku model (4.6+) debut at a score of at least 1420 by June 30, 2027?**: Yes, 85%; settles Thursday, 1 July 2027 (polymarket). [Market](https://polymarket.com/event/next-claude-haiku-model-4pt6-text-arena-debut-20260930)
- **Will the next Claude Haiku model (4.6+) debut at a score of at least 1460 by June 30, 2027?**: No, 78%; settles Thursday, 1 July 2027 (polymarket). [Market](https://polymarket.com/event/next-claude-haiku-model-4pt6-text-arena-debut-20260930)
- **Will the next Claude Haiku model (4.6+) debut at a score of at least 1440 by June 30, 2027?**: Yes, 67%; settles Thursday, 1 July 2027 (polymarket). [Market](https://polymarket.com/event/next-claude-haiku-model-4pt6-text-arena-debut-20260930)
- **Will Meta's Watermelon model debut with a Humanity's Last Exam score of 50% or higher?**: No, 96%; settles Wednesday, 30 June 2027 (polymarket). [Market](https://polymarket.com/event/metas-watermelon-model-humanitys-last-exam-debut)
- **AI bubble burst by June 30, 2027?**: No, 82%; settles Thursday, 1 July 2027 (polymarket). [Market](https://polymarket.com/event/ai-bubble-burst-by)
- **Will Moonshot AI be added to the Chinese Military Companies list by June 30, 2027?**: No, 84%; settles Thursday, 1 July 2027 (polymarket). [Market](https://polymarket.com/event/who-will-be-added-to-chinese-military-companies-list-by-june-30-2027-20260706225129414)