whatmodel.sh / Gemini 4 Argon (limited preview)
Gemini 4 Argon (limited preview)
↗ share
+ contribute
Gemini 4 Argon (limited preview) is a closed-weights language model from Google, released September 2026. Licence: proprietary.
Limited preview · broad public access is not yet available. Access announcement ↗
Catalogue identity notes
Exact Arena entry 'gemini-4-argon-high' in the October 2026 pinned dataset. Comparison identity; no public run route inferred from leaderboard inclusion. Announced September 30, 2026; initially limited to trusted cyber defenders through Fairwind. Broad developer access and the announced introductory pricing are prospective, so no generally available API is listed. Arena evaluated High reasoning.
Model information sources
$ how it compares
The LLMs ranked either side of Gemini 4 Argon (limited preview), out of 828 with a score, then more from Google.
| # | model | score | where | ||
|---|---|---|---|---|---|
| 7 |  GPT-6 Sol | 83±1 | api $4.00/M | ||
| 8 | GPT-5.5 Pro | 82±1 | api $67.50/M | ||
| 9 | Gemini 4 Argon (limited preview) | 82±1 | limited preview | ||
| 10 | GPT-5.6 Sol | 82±1 | api $8.00/M | ||
| 11 |  Claude Fable 5 | 81±1 | api $20.00/M | ||
| more from Google | |||||
| 23 | Gemini 3.8 Flash | 77±1 | api $1.50/M | ||
| 32 | Gemini 3.7 Flash | 76±1 | api $1.50/M | ||
| 39 | Gemini 3.1 Pro (Preview) | 75±1 | api $4.50/M | ||
$ what it's good at
agent tasks▁▁▃▆█▆▅▅▅▄▃▁ #9 of 300 * · 43% of its score · 3 tests
Beats GPT-5.5 Pro here, though they rank higher overall. Trails Claude Fable 5 here, though they rank lower overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| Arena Agent (October 2026 aggregate IPS) | 0.0757 IPS ✎ | ▁▁▁▄▂▃▇█▂▃▁▃ beats 41/49 | 37% |
| DeepSWE v1.1 (resolved) lab | 77.9% ✎ | ▂····▁▁▁▆██▆ beats 28/28 | 5.6% |
| Terminal-Bench 4.0 lab | 57.4% ✎ | ▁·▁··▁▇▇▅██▃ beats 25/30 | — |
coding▁▁▁▃▆██▇▆▄▃▁ #8 of 458 * · 38% of its score · 2 tests
| test | result | vs the others in its run | counts |
|---|---|---|---|
| Arena WebDev (Code Arena) | 1679.5 ✎ | ▁▂▄▃▅█▄▇▆▄▁▁ beats 112/119 | 33% |
| DeepSWE v1.1 (resolved) lab | 77.9% ✎ | ▂····▁▁▁▆██▆ beats 28/28 | 5.6% |
chat & writing▁▁▁▃▅▄▅█▆▆▇▃ #1 of 393 * · 18% of its score · 1 test
Beats GPT-6 Astra, Claude Opus 5.5, GPT-6.1 Sol here, though they rank higher overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| LMArena (text) | 1525.2 ✎ | ▁▁▁▂▃▃▃▆▆▇█▃ beats 379/379 | 18% |
$ where to run it
No verified API or download recorded yet · add one
$ how the score was worked out · LMArena leaderboard dataset 88%, Google DeepMind Gemini model… 11%
Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works
| source | results | share of the score |
|---|---|---|
 LMArena leaderboard dataset | 3 | 88% |
 Google DeepMind Gemini model page and model cards (lab-reported scores) · lab-reported | 2 | 11% |
| prior · pull toward the middle | 1% | |
| weighted average | = 81.8 |
not counted (1)
| Google DeepMind Gemini model page and model cards (lab-reported scores) | Terminal-Bench 4.0 [correct] | 57.4% | fewer than three models have this test yet, or this source ran it for this model alone |
$ improve this page
Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.
data updated · sources and licences · json