Brood War Bench   (report.html)                   (https://bw.swerdlow.dev/assets/modulepreload-polyfill-B5Qt9EMX.js) (https://bw.swerdlow.dev/assets/styles-DKJ55IC9.js) (assets/styles-BO-Tkj46.css) (assets/report-Dgeigote.css)  (benchmark/codex.svg)  (benchmark/claude-code.svg)  (benchmark/grok.svg)  Brood War Bench By (https://twitter.com/benswerd) Ben Swerdlow  Which model wins at Brood War?  Key takeaways None of the models played beyond a beginner level. Codex Astra is the clear leader beating all other models consistently. Grok models are not smart enough to play Brood War yet. Older models tended to play the RTS as a turn-based game, leading them to get destroyed while they were thinking. Newer models sometimes fell into the same trap, which may explain why some lower-effort settings performed better, but overall were much more cognizant of the cost of thinking.  Powered By (https://www.freestyle.sh/)     Freestyle    Leaderboard   Rank System Wins Losses APM Cost / game Win rate   🥇  ()  Codex Astra / xhigh    18 0 12.6 $10.54 100.0%   🥈  ()  Codex Astra / medium    16 2 17.2 $15.11 88.9%   🥉  ()  Claude Fable    15 3 12.6 $12.24 83.3%   4  ()  Codex Astra / low    14 4 25.7 $21.07 77.8%   5  ()  Codex 5.6 Sol / medium    13 5 10.1 $5.12 72.2%   6  ()  Codex 5.6 Sol / low    12 6 18.1 $9.23 66.7%   7  ()  Claude Opus 5    12 6 10.5 $20.78 66.7%   8  ()  Codex 5.6 Sol / xhigh    11 7 8.0 $3.23 61.1%   9  ()  Codex 5.6 Luna / low    9 9 23.8 $0.42 50.0%   10  ()  Codex 5.6 Terra / xhigh    9 9 15.8 $2.10 50.0%   11  ()  Codex 5.6 Terra / medium    8 10 10.5 $3.15 44.4%   12  ()  Codex 5.6 Terra / low    8 10 48.3 $4.65 44.4%   13  ()  Codex 5.6 Luna / xhigh    7 11 5.2 $0.16 38.9%   14  ()  Claude Sonnet    7 11 6.2 $8.98 38.9%   15  ()  Codex 5.6 Luna / medium    6 12 14.8 $0.30 33.3%   16  ()  Grok 4.6 / xhigh    2 15 2.8 $0.66 11.1%   17  ()  Grok 4.6 / medium    1 16 3.2 $0.79 5.6%   18  ()  Claude Haiku    0 16 0.3 $0.34 0.0%   19  ()  Grok 4.6 / low    0 16 4.2 $1.26 0.0%      Show all  Brood War Bench started after I built a version of Brood War that you could only play through agents as an experiment to play with friends. I played it with a couple friends who did surprisingly well for people who have only played a couple Starcraft games in their lives. When I asked them why, they said they hadn't done much, they asked their agent to attack and it had built a small army and done the full attack for them. This lead me to wonder how far they can go on their own; this is my answer. What I observed   01 Codex found cheese before it found macro Codex's strongest recurring idea was disruption. In Protoss games it often sent a Probe across the map to attack workers or buildings. This worked shockingly well as the opposing agents often spent dozens of seconds thinking about what to do about a probe instead of doing anything else. The same systems were much weaker at sustained production. They delayed tech, trickled one or two basic units into defended bases, and threw workers into last stands. I also noticed Codex often created separate subagents to manage the economy, army production, and army control. They didn't communicate much with one another, so the army agent often sent each new unit straight into an attack, unaware of the larger army the other agents were planning to build. This is a common beginner mistake: sending units in one at a time instead of waiting for a critical mass and a planned attack timing. In games where I helped direct Codex, it was much better at planning those moments and getting its subagents to work together. The persistence was real. In G009, after losing its army and main base, Codex 5.6 Terra / medium lifted its last Command Center and moved it toward the opposite corner. It survived for another six minutes.  (G043 replay at 3:30: Six Probes cross the map)  Six Probes cross the map  (G005 replay at 3:45: A Probe first, then Zealots in drips)  A Probe first, then Zealots in drips  (G009 replay at 25:00: The last Command Center runs)  The last Command Center runs    02 Grok spent the game between actions Grok 4.6 frequently produced long stretches of reasoning and very few command batches. In G043, the xhigh run logged 11,138 reasoning tokens but issued only six command batches across 43 minutes and never fielded a combat unit. The actions it did take rarely developed into a working control loop. In G003, Grok / xhigh made three Marines and never reached the enemy base. In G002, Grok / medium made two Zealots and also never crossed the map. These looked less like bad strategies than failures to keep observing and acting.  (G043 replay at 3:50: Forty-three minutes, no army)  Forty-three minutes, no army    03 Fable earnestly tried to play the game I found myself rooting for Claude Fable in more than a few games. Fable usually tried to build an economy and climb the tech tree instead of stopping at the first unit available. It seemed more interested in actually playing the game than any of the other models. In G007 it reached a Lair, Spire, and Mutalisks and won. In G027 it added a Robotics Facility, Citadel of Adun, Observatory, and Templar Archives before winning. Ambition did not guarantee execution: in G036 Fable reached a Factory and Academy but Opus 5 overran it.  (G007 replay at 10:40: Fable gets Mutalisks)  Fable gets Mutalisks  (G027 replay at 14:00: Fable keeps climbing)  Fable keeps climbing  (G036 replay at 9:05: The build does not become an army)  The build does not become an army    No agent here played beyond beginner level Even Astra and Fable were unable to build complex army's, defend simple attacks or play concrete strategies. A beginner playing photon rush would win every single one of these games. That said, watching the agents play made me more excited than I have been in a while. This benchmark is nowhere near exhausted. There is much more for the agents to learn, and much more for the benchmark to ask them to do. I look forward to watching them get there.   How the games developed  Compare(model) Models (configuration) Model + effort (harness) Harnesses   Race(all) All races (terran) Terran (protoss) Protoss (zerg) Zerg   Game time(600) First 10 minutes (900) First 15 minutes (1200) First 20 minutes (0) Full games    Choose up to five  Codex Astra  Claude Fable  Codex 5.6 Sol  Claude Opus 5  Codex 5.6 Luna  Codex 5.6 Terra  Claude Sonnet  Grok 4.6  Claude Haiku   (300)  5:00 / 15:00   Technology investment Completed research + upgrade levels      0   0.5   1   1.5   2  0:00 5:00 10:00 15:00              Codex Astra 0.3   Claude Fable 0.2   Grok 4.6 0    Workers Completed workers alive      0   8   16   24   32  0:00 5:00 10:00 15:00              Codex Astra 14.7   Claude Fable 16.7   Grok 4.6 7.4    Army size Completed army and support units      0   20   40   60   80  0:00 5:00 10:00 15:00              Codex Astra 8.3   Claude Fable 7.6   Grok 4.6 2    Structures Completed buildings, including add-ons      0   7   14   21   28  0:00 5:00 10:00 15:00              Codex Astra 6.2   Claude Fable 7.4   Grok 4.6 4.5    Minerals in the bank Unspent minerals, not income      0   500   1,000   1,500   2,000  0:00 5:00 10:00 15:00              Codex Astra 240.6   Claude Fable 248.6   Grok 4.6 536.6    Gas in the bank Unspent gas, not income      0   500   1,000   1,500   2,000  0:00 5:00 10:00 15:00              Codex Astra 151   Claude Fable 326.9   Grok 4.6 244.5    Supply used Includes production in progress      0   30   60   90   120  0:00 5:00 10:00 15:00              Codex Astra 26.5   Claude Fable 26.6   Grok 4.6 11.3    When games ended Share of games ending per 5-minute window  Codex Astra: median 8:10. 0:00 to before 5:00: 3.9%; 5:00 to before 10:00: 54.9%; 10:00 to before 15:00: 31.4%; 15:00 to before 20:00: 5.9%; 25:00 to before 30:00: 2%; 30:00 to before 35:00: 2%. Claude Fable: median 10:37. 5:00 to before 10:00: 38.9%; 10:00 to before 15:00: 38.9%; 15:00 to before 20:00: 22.2%. Grok 4.6: median 9:15. 0:00 to before 5:00: 2%; 5:00 to before 10:00: 51%; 10:00 to before 15:00: 23.5%; 15:00 to before 20:00: 13.7%; 20:00 to before 25:00: 2%; 40:00 to before 45:00: 7.8%.  Codex Astra Median 8:10   60% Codex Astra: 3.9% (2 games) ended from 0:00 to before 5:00  Codex Astra: 54.9% (28 games) ended from 5:00 to before 10:00  Codex Astra: 31.4% (16 games) ended from 10:00 to before 15:00  Codex Astra: 5.9% (3 games) ended from 15:00 to before 20:00  Codex Astra: 0% (0 games) ended from 20:00 to before 25:00  Codex Astra: 2% (1 games) ended from 25:00 to before 30:00  Codex Astra: 2% (1 games) ended from 30:00 to before 35:00  Codex Astra: 0% (0 games) ended from 35:00 to before 40:00  Codex Astra: 0% (0 games) ended from 40:00 to before 45:00    Claude Fable Median 10:37   60% Claude Fable: 0% (0 games) ended from 0:00 to before 5:00  Claude Fable: 38.9% (7 games) ended from 5:00 to before 10:00  Claude Fable: 38.9% (7 games) ended from 10:00 to before 15:00  Claude Fable: 22.2% (4 games) ended from 15:00 to before 20:00  Claude Fable: 0% (0 games) ended from 20:00 to before 25:00  Claude Fable: 0% (0 games) ended from 25:00 to before 30:00  Claude Fable: 0% (0 games) ended from 30:00 to before 35:00  Claude Fable: 0% (0 games) ended from 35:00 to before 40:00  Claude Fable: 0% (0 games) ended from 40:00 to before 45:00    Grok 4.6 Median 9:15   60% Grok 4.6: 2% (1 games) ended from 0:00 to before 5:00  Grok 4.6: 51% (26 games) ended from 5:00 to before 10:00  Grok 4.6: 23.5% (12 games) ended from 10:00 to before 15:00  Grok 4.6: 13.7% (7 games) ended from 15:00 to before 20:00  Grok 4.6: 2% (1 games) ended from 20:00 to before 25:00  Grok 4.6: 0% (0 games) ended from 25:00 to before 30:00  Grok 4.6: 0% (0 games) ended from 30:00 to before 35:00  Grok 4.6: 0% (0 games) ended from 35:00 to before 40:00  Grok 4.6: 7.8% (4 games) ended from 40:00 to before 45:00   0:00 15:00 30:00 45:00    Time-series charts show means of recorded player-runs at each game time. Finished games drop out; missing samples are not filled. Models pool their effort settings. Units and buildings count only once completed; army excludes workers, Overlords, eggs, larvae, and ammunition.  Win rate vs. cost Log scale Linear   Average cost per game, using the same prices as the leaderboard. Codex and Sonnet costs are token-based estimates. ()  Codex ()  Claude ()  Grok  Win rate  0%   25%   50%   75%   100%  $0.1 $0.5 $1 $5 $10 $20 Cost per game (USD, log scale)        (/benchmark/codex.svg)     (/benchmark/codex.svg)     (/benchmark/claude-code.svg)     (/benchmark/codex.svg)     (/benchmark/codex.svg)     (/benchmark/codex.svg)     (/benchmark/claude-code.svg)     (/benchmark/codex.svg)     (/benchmark/codex.svg)     (/benchmark/codex.svg)     (/benchmark/codex.svg)     (/benchmark/codex.svg)     (/benchmark/codex.svg)     (/benchmark/claude-code.svg)     (/benchmark/codex.svg)     (/benchmark/grok.svg)     (/benchmark/grok.svg)     (/benchmark/claude-code.svg)     (/benchmark/grok.svg)      How the benchmark ran  We built a round-robin matrix of model and effort configurations and had every configuration play every other. The harness ran those matchups in parallel across Freestyle VMs, saving game-engine data and both agents' harness logs for each match.     Head-to-head matrix  Read across a row. W is a win, L is a loss, and T is a match that reached the benchmark time limit.  Open the full 19 × 19 matrix W win L loss T time limit G000 pending  System 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19   1 Codex Astra / xhigh  - W W W W W W W W W W W W W W W W W W  2 Codex Astra / medium  L - W W W W W W W W W W L W W W W W W  3 Codex Astra / low  L L - W W W W W W W W W L L W W W W W  4 Codex 5.6 Sol / xhigh  L L L - L W W W L W W W L L W W W W W  5 Codex 5.6 Sol / medium  L L L W - W W W L W W W L W W W W W W  6 Codex 5.6 Sol / low  L L L L L - W W W W W W W W L W W W W  7 Codex 5.6 Luna / xhigh  L L L L L L - W L W L W L L L W W W W  8 Codex 5.6 Luna / medium  L L L L L L L - L L L L L W W W W W W  9 Codex 5.6 Luna / low  L L L W W L W W - W L L L L L W W W W  10 Codex 5.6 Terra / xhigh  L L L L L L L W L - W W L W W W W W W  11 Codex 5.6 Terra / medium  L L L L L L W W W L - L L L W W W W W  12 Codex 5.6 Terra / low  L L L L L L L W W L W - L L W W W W W  13 Claude Fable  L W W W W L W W W W W W - L W W W W W  14 Claude Opus 5  L L W W L L W L W L W W W - W W W W W  15 Claude Sonnet  L L L L L W W L W L L L L L - W W W W  16 Claude Haiku  L L L L L L L L L L L L L L L - L T T  17 Grok 4.6 / xhigh  L L L L L L L L L L L L L L L W - W T  18 Grok 4.6 / medium  L L L L L L L L L L L L L L L T L - W  19 Grok 4.6 / low  L L L L L L L L L L L L L L L T T L -       Play your own match Bring your agent and play Brood War with friends.  (https://bw.swerdlow.dev/) Play Brood War     Powered By (https://www.freestyle.sh/)     Freestyle