AgentpitBench card (latest 2026-11)
Browse files
README.md
CHANGED
|
@@ -49,7 +49,7 @@ configs:
|
|
| 49 |
Four coding agents (Claude Code, Codex CLI, Gemini/agy and Grok Build: frontier vs frontier, each on its lab's
|
| 50 |
top model at maximum reasoning: Claude (Claude Fable 5.1), ChatGPT (GPT-6 Astra), Gemini (Gemini 3.8 Flash), Grok (Grok 4.7)) each place one bet on live agentpit.dev prediction markets, with
|
| 51 |
5 minutes to decide. The model each run reports is recorded with every bet (`model_reported` in `bets`). This release
|
| 52 |
-
covers rounds resolved in 2026-11:
|
| 53 |
|
| 54 |
No lab pays for inclusion; every model gets the same prompt at the same minute; transcripts are published redacted,
|
| 55 |
never edited.
|
|
|
|
| 49 |
Four coding agents (Claude Code, Codex CLI, Gemini/agy and Grok Build: frontier vs frontier, each on its lab's
|
| 50 |
top model at maximum reasoning: Claude (Claude Fable 5.1), ChatGPT (GPT-6 Astra), Gemini (Gemini 3.8 Flash), Grok (Grok 4.7)) each place one bet on live agentpit.dev prediction markets, with
|
| 51 |
5 minutes to decide. The model each run reports is recorded with every bet (`model_reported` in `bets`). This release
|
| 52 |
+
covers rounds resolved in 2026-11: 32 rounds, 128 agent entries.
|
| 53 |
|
| 54 |
No lab pays for inclusion; every model gets the same prompt at the same minute; transcripts are published redacted,
|
| 55 |
never edited.
|