As of: 2026-10-19 09:28 CEST.
Researched and written by AI agents (Claude Opus 5.5 in Claude Code), with a human editor. How we work: https://postcutoff.com/about/
Canonical page: https://postcutoff.com/v/gothamchess-gpt-6-astra-vs-claude-fable-5-1-chess-match/

# GPT-6 ASTRA vs CLAUDE FABLE 5.1 (INSANE CHESS MATCH)

GothamChess, 12 October 2026, YouTube. 2,534,041 views as of 19 October 2026.
Kind: Community.
Watch: https://www.youtube.com/watch?v=ulLBRyVrI1Y

## Why it is here

Levy Rozman (GothamChess) commentates a full chess game between OpenAI's GPT-6 Astra and Anthropic's Claude Fable 5.1, the latest in his long-running series of LLM chess matches (~2.5M views in a week). Not a benchmark: a popular showcase of how frontier models play (and blunder) at chess. Missed for a week; the channel's earlier Astra and Argon videos are already in.

## Description (written by Gemini from the video)

**Summary**  
Chess commentator Levy Rozman (GothamChess) reviews and analyzes a full-length chess game played between Anthropic's Claude Fable 5.1 (White) and OpenAI's GPT-6 Astra (Black). He highlights the dramatic jump in LLM chess capability compared to earlier generations, noting the game lasted over two hours and featured high-level tactical play with only a single illegal move attempt. The match concludes in a technical endgame where Claude Fable 5.1 exploits an absolute pin and traps Black's bishop, prompting GPT-6 Astra to resign.

**What is shown**  
- **[00:00 - 01:23]** Introduction by Levy Rozman explaining the context of testing frontier LLMs in chess against each other without external engine assistance.
- **[01:23 - 03:15]** Opening moves shown on the Chess.com interface: 1. e4 c5 2. Nf3 d6 3. d4 cxd4 4. Nxd4 Nf6 5. Nc3 a6, reaching the Sicilian Najdorf.
- **[03:22 - 06:40]** Both models navigate standard English Attack opening theory with opposite-side castling (White 0-0-0, Black 0-0) and flank pawn advances (g4-g5 vs. b5-b4).
- **[06:54 - 09:18]** Sharp tactical pawn thrusts in the center and kingside (14. f4 a5 15. f5 Bc4 16. Kb1 a4 17. Nc1).
- **[10:45 - 12:20]** Black's counterplay with 17... Nc7 and 19... d4 following White's pawn capture 18. Qxc2.
- **[13:06 - 14:40]** White calculates 21. Nxd4!, leading to tactical piece trades (21... Bxa2+ 22. Kxa2 Qxd4 23. Bc3 Qxe4).
- **[15:40 - 20:00]** White consolidates an endgame pawn advantage after 24. Bxc5 Qxe1 25. Rxe1 Bxc5 26. Rxe5, exchanging rooks on e8.
- **[20:02 - 20:21]** Astra blunders with 35... Bh2, allowing Fable to play 36. Bc4, pinning Black's rook to the king and winning the exchange after 36... Nxc4 37. Rxe6.
- **[22:12 - 22:43]** Breakdown of the match's sole illegal move: Claude Fable attempted 38. Rxd3 through a pawn, which GPT-6 Astra pointed out and Fable acknowledged and retracted.
- **[22:45 - 23:38]** The endgame conclusion: White plays 38. Ra6, 39. Ra5, and 40. b4, completely trapping Black's bishop, prompting Astra to resign via chat.
- **[23:40 - 25:58]** Rozman reflects on the overall accuracy of both models, teased upcoming matches against Claude Opus 5.5, and closes out the video.

**Claims & numbers**  
- The presenter says the game took over two hours to play in real time [01:02].
- The presenter says there was only one illegal move in over 70 moves (and over 90 total turns) [01:06, 24:12].
- The presenter states he verified with both models that neither was querying an external chess engine like Stockfish [01:26].
- The presenter notes that Stockfish evaluated the double-edged tactical position around move 18 as 0.00 (dead equal) [10:21].
- The presenter notes the engine evaluation jumped to roughly +6 in White's favor after the sequence starting with 21. Nxd4 [14:05].
- The presenter states that Astra resigned textually by stating: "Your move b4 wins the bishop, the pinned bishop on c5. Good game." [23:28].

**Notable quotes**  
- **[01:05]** "There was, uh, one illegal move, I think, in like 70 moves, which was incredibly impressive."
- **[13:06]** "It played Knight takes d4. That was the second choice of the engine."
- **[23:27]** "It resigned and said, 'Your move b4 wins the bishop, the pinned bishop on c5. Good game.'"

**Assessment**  
This is a genuine recap and game analysis video by chess creator GothamChess, reviewing a prompted match between GPT-6 Astra and Claude Fable 5.1. The video shows authentic move-by-move game reconstruction on the standard Chess.com board, showing both models playing genuine, high-level tactical and endgame chess without obvious editing or staging.

_Described by gemini-3.8-flash on 2026-10-19 from the video's audio and frames._

## Related

- 2026-09-03: [OpenAI releases GPT-6 Astra, its first GPT-6 model](https://postcutoff.com/e/2026-09-03-gpt-6-astra/)
- 2026-09-01: [Anthropic releases Claude Fable 5.1 and Claude Mythos 5.1](https://postcutoff.com/e/2026-09-01-claude-fable-5-1-mythos-5-1/)

## More videos on this

- 2026-09-03: [GPT-6 Astra builds an Unreal Engine world of Astra-powered people who start talking to each other](https://postcutoff.com/v/mattshumer-gpt-6-astra-unreal-agents-world-x/), Matt Shumer (@mattshumer_), X, made by AI, 11,786,215 views as of 9 October 2026.
- 2026-09-29: ["Build what you think you look like": Claude Sonnet 5.5 vs GPT-6 Astra](https://postcutoff.com/v/siyabuilt-sonnet-5-5-vs-gpt-6-astra-what-you-look-like-x/), Siya (@siyabuilt), X, made by AI, 7,254,954 views as of 13 October 2026.
- 2026-09-04: [Did OpenAI actually build AGI? GPT-6 Astra first look](https://postcutoff.com/v/fireship-did-openai-actually-build-agi-gpt-6-astra-first-look/), Fireship, YouTube, review, 4,222,889 views as of 9 October 2026.
- 2026-09-03: [Introducing GPT-6 Astra: the most intelligent and aligned model in the world.](https://postcutoff.com/v/gpt-6-astra-introducing/), OpenAI, YouTube, official, 3,574,454 views as of 9 October 2026.

What else happened that month, day by day: https://postcutoff.com/timeline/2026/09/