| Gemini 4 Argon |
All 8 badges in Pokémon Emerald (now 3,722 turns in, still playing) Last verified: Oct 18, 2026 · Last checked: Oct 18, 2026 How was this run measured?
- Model
- Gemini 4 Argon
- Game
- Pokémon Emerald
- Project
- Gemini Plays Pokémon
- Harness
- Gemini Plays Pokémon harness (mGBA emulator)
- Inputs
- emulator frames + game-state analysis
- Tools
- custom subagents, Python sandbox, notepads (published on GitHub)
- Assistance
- not disclosed
- Budget
- not disclosed
- Verification
- Verified Badge count and turn count read from the hub's game-state telemetry; every turn is committed to the public repo with an emulator save state.
- Evidence
- Hub telemetry: run emerald_gemini_4_argon (primary)
Public repo branch emerald_gemini_4_argon (turn-by-turn commits) (primary)
- Unknown
- turn at which the 8th badge was earned; reasoning budget
- Dates
- started Oct 13, 2026 · evidence dated Oct 18, 2026
- Recorded
- published Oct 18, 2026 · discovered by AI vs Games Oct 18, 2026
- Changes
- Last change 2026-10-18: Turns 3707 → 3722
Permanent link · Submit a correction · Share card (PNG) · JSON |
Gemini Plays Pokémon harness (mGBA emulator) emulator frames + game-state analysis |
Verified |
Oct 18, 2026 |
| Gemini 4 Argon |
Entered the Hall of Fame in Pokémon Emerald (beat Champion Wallace) Gemini 4 Argon's whole run so far (3,722 turns, post-game included) is under half the 8,825 turns Gemini 3.8 Flash has used to reach 5 badges on the same hub. Last verified: not independently verified · Last checked: Oct 18, 2026 How was this run measured?
- Model
- Gemini 4 Argon
- Game
- Pokémon Emerald
- Project
- Gemini Plays Pokémon
- Harness
- Gemini Plays Pokémon harness (mGBA emulator)
- Inputs
- emulator frames + game-state analysis
- Tools
- custom subagents, Python sandbox, notepads (published on GitHub)
- Assistance
- not disclosed
- Budget
- not disclosed
- Verification
- Creator-reported Stated in the AI's own run notepad, which logs post-Hall of Fame events. The hub does not expose a Hall of Fame flag, so we have not confirmed it from game state.
- Evidence
- Hub telemetry: run emerald_gemini_4_argon (primary)
Public repo branch emerald_gemini_4_argon (turn-by-turn commits) (primary)
Run notepad: 'Post-Hall of Fame' section (creator)
- Unknown
- turn and date of the Hall of Fame entry
- Dates
- started Oct 13, 2026 · evidence dated Oct 18, 2026
- Recorded
- published Oct 18, 2026 · discovered by AI vs Games Oct 18, 2026
- Changes
- Historical record. Nothing to update unless the source changes.
Permanent link · Submit a correction · JSON |
Gemini Plays Pokémon harness (mGBA emulator) emulator frames + game-state analysis |
Creator-reported |
Oct 18, 2026 |
| Gemini 3.8 Flash |
5 of 8 badges in Pokémon Emerald after 8,825 turns (live) Picked up badge 4 (Flannery, Lavaridge Gym) overnight on Oct 17–18 after a long stay in Lavaridge Town. Last verified: Oct 18, 2026 · Last checked: Oct 18, 2026 How was this run measured?
- Model
- Gemini 3.8 Flash
- Game
- Pokémon Emerald
- Project
- Gemini Plays Pokémon
- Harness
- Gemini Plays Pokémon harness (mGBA emulator)
- Inputs
- emulator frames + game-state analysis
- Tools
- custom subagents, Python sandbox, notepads (published on GitHub)
- Assistance
- not disclosed
- Budget
- not disclosed
- Verification
- Verified Hub game-state telemetry plus turn commits in the public repo.
- Evidence
- Hub telemetry: run emerald_gemini_3_8_flash (primary)
Public repo branch emerald_gemini_3_8_flash (turn-by-turn commits) (primary)
- Dates
- started Oct 13, 2026 · evidence dated Oct 18, 2026
- Recorded
- published Oct 18, 2026 · discovered by AI vs Games Oct 17, 2026
- Changes
- Last change 2026-10-18: Turns 8710 → 8825
Permanent link · Submit a correction · Share card (PNG) · JSON |
Gemini Plays Pokémon harness (mGBA emulator) emulator frames + game-state analysis |
Verified |
Oct 18, 2026 |
| GPT-5.6 Sol |
Became Champion of Pokémon Emerald in 153h51 (8 badges) Last verified: not independently verified · Last checked: Oct 18, 2026 How was this run measured?
- Model
- GPT-5.6 Sol
- Game
- Pokémon Emerald
- Project
- GPT Plays Pokémon
- Harness
- Codex App Server + custom tools
- Inputs
- vision-only
- Tools
- custom tools (not itemised)
- Assistance
- not disclosed
- Budget
- not disclosed
- Verification
- Awaiting verification Sources disagree: BenchmarkList lists 153h51; the creator's later thread says 96h 35m for GPT-5.6 Sol (max). They may be different runs.
- Evidence
- BenchmarkList: Pokémon Autonomous Completion (self-reported) (secondary)
Creator thread (96h 35m figure) (creator)
- Unknown
- which figure belongs to which run
- Dates
- achieved Jul 21, 2026 · evidence dated Jul 21, 2026
- Recorded
- published Jul 21, 2026 · discovered by AI vs Games Oct 17, 2026
- Changes
- Historical record. Nothing to update unless the source changes.
Permanent link · Submit a correction · JSON |
Codex App Server + custom tools vision-only |
Awaiting verification |
Jul 21, 2026 |
| Heatz agent (model not disclosed) Heatz |
1st: reached the first gym (Roxanne) in 40:13 (1,608 actions) Nearly twice as fast as 2nd place. Last verified: Oct 18, 2026 · Last checked: Oct 18, 2026 How was this run measured?
- Model
- Heatz agent (model not disclosed) · agent: Heatz
- Game
- Pokémon Emerald
- Project
- PokéAgent Challenge (NeurIPS 2025, speedrun track)
- Harness
- PokéAgent Challenge Emerald environment (real-time; game keeps running while the agent thinks)
- Inputs
- per team (documented in the paper's appendix)
- Tools
- Scripted Policy Distillation: an LLM writes scripted policies, distilled into a neural net and refined with RL
- Assistance
- none allowed during evaluation runs
- Budget
- scored on wall-clock time and action count
- Verification
- Verified Official results published by the challenge organizers.
- Evidence
- PokéAgent Challenge paper (organizers), Sec. 5.2 & App. C (operator)
PokéAgent Challenge: Speedrunning track (operator)
- Unknown
- underlying model(s) used by the team (not in our source)
- Dates
- evidence dated Mar 16, 2026
- Recorded
- published Mar 16, 2026 · discovered by AI vs Games Oct 18, 2026
- Changes
- Historical record. Nothing to update unless the source changes.
Permanent link · Submit a correction · Share card (PNG) · JSON |
PokéAgent Challenge Emerald environment (real-time; game keeps running while the agent thinks) per team (documented in the paper's appendix) |
Verified |
Mar 16, 2026 |
| Hamburg PokéRunners agent (model not disclosed) Hamburg PokéRunners |
2nd: reached the first gym in 1:14:43 Last verified: Oct 18, 2026 · Last checked: Oct 18, 2026 How was this run measured?
- Model
- Hamburg PokéRunners agent (model not disclosed) · agent: Hamburg PokéRunners
- Game
- Pokémon Emerald
- Project
- PokéAgent Challenge (NeurIPS 2025, speedrun track)
- Harness
- PokéAgent Challenge Emerald environment (real-time; game keeps running while the agent thinks)
- Inputs
- per team (documented in the paper's appendix)
- Tools
- Pure RL: recurrent PPO with milestone-based rewards
- Assistance
- none allowed during evaluation runs
- Budget
- scored on wall-clock time and action count
- Verification
- Verified Official results published by the challenge organizers.
- Evidence
- PokéAgent Challenge paper (organizers), Sec. 5.2 & App. C (operator)
PokéAgent Challenge: Speedrunning track (operator)
- Unknown
- underlying model(s) used by the team (not in our source)
- Dates
- evidence dated Mar 16, 2026
- Recorded
- published Mar 16, 2026 · discovered by AI vs Games Oct 18, 2026
- Changes
- Historical record. Nothing to update unless the source changes.
Permanent link · Submit a correction · JSON |
PokéAgent Challenge Emerald environment (real-time; game keeps running while the agent thinks) per team (documented in the paper's appendix) |
Verified |
Mar 16, 2026 |
| anthonys agent (model not disclosed) anthonys |
3rd: reached the first gym in 1:29:17 Last verified: Oct 18, 2026 · Last checked: Oct 18, 2026 How was this run measured?
- Model
- anthonys agent (model not disclosed) · agent: anthonys
- Game
- Pokémon Emerald
- Project
- PokéAgent Challenge (NeurIPS 2025, speedrun track)
- Harness
- PokéAgent Challenge Emerald environment (real-time; game keeps running while the agent thinks)
- Inputs
- per team (documented in the paper's appendix)
- Tools
- Tool-calling VLM policy (LLM harness)
- Assistance
- none allowed during evaluation runs
- Budget
- scored on wall-clock time and action count
- Verification
- Verified Official results published by the challenge organizers.
- Evidence
- PokéAgent Challenge paper (organizers), Sec. 5.2 & App. C (operator)
PokéAgent Challenge: Speedrunning track (operator)
- Unknown
- underlying model(s) used by the team (not in our source)
- Dates
- evidence dated Mar 16, 2026
- Recorded
- published Mar 16, 2026 · discovered by AI vs Games Oct 18, 2026
- Changes
- Historical record. Nothing to update unless the source changes.
Permanent link · Submit a correction · JSON |
PokéAgent Challenge Emerald environment (real-time; game keeps running while the agent thinks) per team (documented in the paper's appendix) |
Verified |
Mar 16, 2026 |
| Deepest agent (model not disclosed) Deepest |
Judge's Choice: fewest actions to the first gym (649), 5th by time Slowest clock, fewest button presses: a good example of why we never merge metrics. Last verified: Oct 18, 2026 · Last checked: Oct 18, 2026 How was this run measured?
- Model
- Deepest agent (model not disclosed) · agent: Deepest
- Game
- Pokémon Emerald
- Project
- PokéAgent Challenge (NeurIPS 2025, speedrun track)
- Harness
- PokéAgent Challenge Emerald environment (real-time; game keeps running while the agent thinks)
- Inputs
- per team (documented in the paper's appendix)
- Tools
- LLM harness
- Assistance
- none allowed during evaluation runs
- Budget
- scored on wall-clock time and action count
- Verification
- Verified Official results published by the challenge organizers.
- Evidence
- PokéAgent Challenge paper (organizers), Sec. 5.2 & App. C (operator)
PokéAgent Challenge: Speedrunning track (operator)
- Unknown
- underlying model(s) used by the team (not in our source)
- Dates
- evidence dated Mar 16, 2026
- Recorded
- published Mar 16, 2026 · discovered by AI vs Games Oct 18, 2026
- Changes
- Historical record. Nothing to update unless the source changes.
Permanent link · Submit a correction · JSON |
PokéAgent Challenge Emerald environment (real-time; game keeps running while the agent thinks) per team (documented in the paper's appendix) |
Verified |
Mar 16, 2026 |
