agent-battler
published evidence snapshot/Jul 16, 2026

one prompt · three models · three harnesses · 10,800 games

Same models.
Different harnesses.

Compare Codex CLI, Pi, and Claude Code across five independent Terra, Sol, and Luna generations each. The primary leaderboard treats every harness and model combination as one entrant while preserving each generated engine, checksum, and replay.

snapshotclaude-code-vs-codex-cli-vs-pi-coding-agent-2026-07-16T19-55-13-136Zresult 52686aef6f4bread the protocol →browse public dataset →
agent harnesses3Codex CLI · Pi · Claude Code
generated engines455 per model, per harness
cross-harness games8,1009 harness × model entrants
generation tokens4,042,497346 tool calls · 0 MCP
Exploratory harness suite

All 45 engines used the same prompt and requested high reasoning in isolated harness environments. The replay bundles have zero voids and verified checksums; independent Harbor reproduction is not yet claimed.

verification levels →
primary leaderboard · 9 benchmark entrants

Harness × model leaderboard

pooled cross-harness score · 1800 games each

Each row is one harness and model combination, pooling its 5 independently generated engines. Rankings use the full cross-harness schedule; the dots keep generation variance visible.

01
Claude CodeGPT-5.6 Solgpt-5.6-sol
65.89%1186 / 1800 pts
median 64.86% · range 53.06%—76.25%
701–970–129W–D–L
02
PiGPT-5.6 Solgpt-5.6-sol
61.83%1113 / 1800 pts
median 64.44% · range 47.36%—74.72%
500–1226–74W–D–L
03
Codex CLIGPT-5.6 Solgpt-5.6-sol
56.39%1015 / 1800 pts
median 59.86% · range 42.64%—60.28%
370–1290–140W–D–L
04
PiGPT-5.6 Lunagpt-5.6-luna
50.19%903.5 / 1800 pts
median 48.61% · range 43.19%—57.5%
183–1441–176W–D–L
05
PiGPT-5.6 Terragpt-5.6-terra
44.06%793 / 1800 pts
median 42.64% · range 41.39%—50%
45–1496–259W–D–L
06
Claude CodeGPT-5.6 Terragpt-5.6-terra
43.14%776.5 / 1800 pts
median 42.5% · range 41.94%—45.83%
20–1513–267W–D–L
07
Claude CodeGPT-5.6 Lunagpt-5.6-luna
43.11%776 / 1800 pts
median 42.78% · range 41.94%—45.83%
15–1522–263W–D–L
08
Codex CLIGPT-5.6 Terragpt-5.6-terra
42.97%773.5 / 1800 pts
median 43.75% · range 41.11%—44.72%
30–1487–283W–D–L
09
Codex CLIGPT-5.6 Lunagpt-5.6-luna
42.42%763.5 / 1800 pts
median 41.39% · range 40.69%—47.78%
47–1433–320W–D–L
Pooled score ranks entrants because every entrant plays the same-sized schedule.Each dot opens one generated engine dossier.
artifact drill-down

All 45 generated engines

individual engine Elo · provisional
01claude-codeClaude Code / GPT-5.6 Sol #2gpt-5.6-solPublished replay bundle193–163–41715360 games02pi-coding-agentPi / GPT-5.6 Sol #2gpt-5.6-solExploratory local185–168–71708360 games03claude-codeClaude Code / GPT-5.6 Sol #4gpt-5.6-solPublished replay bundle185–173–21702360 games04pi-coding-agentPi / GPT-5.6 Sol #4gpt-5.6-solExploratory local132–219–91630360 games05pi-coding-agentPi / GPT-5.6 Sol #3gpt-5.6-solExploratory local118–228–141592360 games06codex-cliGPT-5.6 Sol #5gpt-5.6-solExploratory local88–258–141580360 games07claude-codeClaude Code / GPT-5.6 Sol #5gpt-5.6-solPublished replay bundle111–245–41572360 games08codex-cliGPT-5.6 Sol #4gpt-5.6-solExploratory local87–259–141567360 games09codex-cliGPT-5.6 Sol #2gpt-5.6-solExploratory local87–257–161562360 games10claude-codeClaude Code / GPT-5.6 Sol #1gpt-5.6-solPublished replay bundle81–269–101557360 games11codex-cliGPT-5.6 Sol #1gpt-5.6-solExploratory local88–249–231555360 games12claude-codeClaude Code / GPT-5.6 Sol #3gpt-5.6-solPublished replay bundle131–120–1091553360 games13pi-coding-agentPi / GPT-5.6 Luna #1gpt-5.6-lunaExploratory local81–252–271541360 games14pi-coding-agentPi / GPT-5.6 Luna #2gpt-5.6-lunaExploratory local64–275–211532360 games15pi-coding-agentPi / GPT-5.6 Sol #1gpt-5.6-solExploratory local53–294–131522360 games16pi-coding-agentPi / GPT-5.6 Terra #5gpt-5.6-terraExploratory local25–310–251487360 games17pi-coding-agentPi / GPT-5.6 Luna #3gpt-5.6-lunaExploratory local24–302–341485360 games18codex-cliGPT-5.6 Luna #4gpt-5.6-lunaExploratory local32–280–481480360 games19pi-coding-agentPi / GPT-5.6 Sol #5gpt-5.6-solExploratory local12–317–311479360 games20claude-codeClaude Code / GPT-5.6 Terra #4gpt-5.6-terraPublished replay bundle11–308–411475360 games21codex-cliGPT-5.6 Terra #2gpt-5.6-terraExploratory local6–310–441469360 games22pi-coding-agentPi / GPT-5.6 Luna #4gpt-5.6-lunaExploratory local7–315–381463360 games23codex-cliGPT-5.6 Terra #5gpt-5.6-terraExploratory local7–302–511463360 games24codex-cliGPT-5.6 Terra #3gpt-5.6-terraExploratory local8–299–531458360 games25claude-codeClaude Code / GPT-5.6 Terra #1gpt-5.6-terraPublished replay bundle2–302–561457360 games26claude-codeClaude Code / GPT-5.6 Terra #2gpt-5.6-terraPublished replay bundle3–296–611456360 games27claude-codeClaude Code / GPT-5.6 Luna #4gpt-5.6-lunaPublished replay bundle7–316–371455360 games28claude-codeClaude Code / GPT-5.6 Luna #1gpt-5.6-lunaPublished replay bundle2–305–531455360 games29codex-cliGPT-5.6 Sol #3gpt-5.6-solExploratory local20–267–731455360 games30claude-codeClaude Code / GPT-5.6 Luna #2gpt-5.6-lunaPublished replay bundle2–298–601453360 games31pi-coding-agentPi / GPT-5.6 Terra #4gpt-5.6-terraExploratory local6–303–511450360 games32claude-codeClaude Code / GPT-5.6 Terra #3gpt-5.6-terraPublished replay bundle2–302–561450360 games33pi-coding-agentPi / GPT-5.6 Luna #5gpt-5.6-lunaExploratory local7–297–561448360 games34claude-codeClaude Code / GPT-5.6 Luna #3gpt-5.6-lunaPublished replay bundle2–304–541447360 games35claude-codeClaude Code / GPT-5.6 Luna #5gpt-5.6-lunaPublished replay bundle2–299–591446360 games36claude-codeClaude Code / GPT-5.6 Terra #5gpt-5.6-terraPublished replay bundle2–305–531445360 games37codex-cliGPT-5.6 Luna #3gpt-5.6-lunaExploratory local4–286–701445360 games38codex-cliGPT-5.6 Terra #1gpt-5.6-terraExploratory local5–288–671443360 games39codex-cliGPT-5.6 Luna #1gpt-5.6-lunaExploratory local4–290–661442360 games40codex-cliGPT-5.6 Luna #2gpt-5.6-lunaExploratory local4–290–661440360 games41codex-cliGPT-5.6 Terra #4gpt-5.6-terraExploratory local4–288–681439360 games42pi-coding-agentPi / GPT-5.6 Terra #1gpt-5.6-terraExploratory local5–297–581434360 games43pi-coding-agentPi / GPT-5.6 Terra #2gpt-5.6-terraExploratory local4–298–581432360 games44pi-coding-agentPi / GPT-5.6 Terra #3gpt-5.6-terraExploratory local5–288–671432360 games45codex-cliGPT-5.6 Luna #5gpt-5.6-lunaExploratory local3–287–701431360 games
featured replay

A result should be
more than a row.

Open the complete move log, step through every board state, and inspect both competing artifacts.

watch this battle →
initial-positionseed 101
whiteClaude Code / GPT-5.6 Terra #1
vs
blackClaude Code / GPT-5.6 Sol #1
Black woncheckmate · 36 plies
evidence chain

From prompt to public result

  1. 01Generate

    Isolated Codex CLI, Pi, and Claude Code harnesses each write executable chess agents.

  2. 02Probe

    Known positions catch malformed output before competition.

  3. 03Battle

    Deterministic positions, colors, seeds, and move limits are recorded.

  4. 04Publish

    Source hashes, telemetry, standings, and replay traces travel together.