Gemini 4 Argon leads 13 of the 19 rows in Google’s launch table outright, ties 1 (CWE-bench v1, with GPT-6 Astra), and trails on 5: FrontierSWE v2, Terminal-bench 4.0, PostTrainBench, Terminal-Bench Science 0.1 and OSWorld-2.0. Since October 13, 2026, you can check them yourself: gemini-4-argon is generally available on the paid Gemini API tier. Independent results are arriving too: Cursor measures 52.8% on CursorBench, “on par with Opus 5.5,” and Scale AI ranks Argon first on Humanity’s Sixth Sense.
Below: the full table with who measured each row, the five losses, the fine print, third-party scoreboards as of October 15, 2026, and how to rerun the rows you care about in Apidog. For the model overview, start with what Gemini 4 Argon is. For the buying decision, see Argon vs GPT-6 Astra vs Claude Opus 5.5.
The full table, with who measured each row
These are Google-reported results from the launch post and the evals methodology PDF. Google computed 10 of the 19 Argon scores itself; the other 9 come from public leaderboards. Rival numbers come from those leaderboards, Google’s runs, or vendor system cards and blog posts, and harnesses differ by row. Bold marks the row leader.
| Benchmark | Gemini 4 Argon | GPT-6 Astra | Claude Fable 5.1 | Claude Opus 5.5 | Who measured it |
|---|---|---|---|---|---|
| Vals Index | 68.9% | 63.1% | 65.8% | 67.0% | Vals AI |
| AutomationBench | 51.3% | 41.4% | 31.4% | 42.5% | Zapier leaderboard (private set) |
| Vals Finance Agent v2 | 65.4% | 53.5% | 58.9% | 58.6% | Vals AI |
| Harvey Legal Agent | 19.6% | 5.4% | 6.7% | 3.8% | Vals AI |
| DeepSWE v1.1 | 77.9% | 74.1% | 67.4% | 74.2% | Argon self-computed; Astra leaderboard; Anthropic system cards |
| FrontierSWE v2 | 55.0% | 65.5% | 56.3% | 62.3% | Proximal leaderboard |
| Vibe Code Bench | 91.9% | 89.6% | 90.3% | 90.3% | Vals AI |
| Terminal-bench 4.0 | 57.4% | 58.2% | 57.9% | 66.4% | Argon self-computed; rivals leaderboard |
| PostTrainBench | 45.3% | 44.3% | 40.2% | 49.3% | Self-computed for all models |
| Terminal-Bench Science 0.1 | 57.6% | 68.1% | 52.6% | 63.3% | Argon self-computed; rivals leaderboard |
| LABBench 2 | 88.8% | 85.4% | 68.6% | 73.1% | Self-computed for all models |
| RiemannBench | 76.0% | 72.0% | 65.6% | 69.6% | Surge leaderboard |
| GraphWalks up to 128K | 99.7% | 98.7% | 91.4% | 90.6% | Self-computed for all models |
| GraphWalks 256K to 1M | 84.2% | 71.8% | 65.0% | 66.8% | Self-computed for all models |
| Agent’s Last Exam | 39.5% | 34.2% | n/r | 38.2% | Argon self-computed; rivals leaderboard |
| OSWorld-2.0 (offline) | 69.2% | 72.6% | n/r | n/r | Argon self-computed; Astra from OpenAI’s blog post |
| Chartography | 71.6% | 71.0% | 46.2% | 66.3% | Surge leaderboard |
| LVBench | 91.7% | 87.5% | 79.7% | 83.7% | Self-computed for all models |
| CWE-bench v1 | 68.0% | 68.0% | 58.0% | 67.0% | Public leaderboard |
n/r = not reported. Grok 4.7, which isn’t in Google’s table, also scores 68% on the CWE-bench leaderboard, so that tie is three-way.
Sorted by source, 9 rows come from public leaderboards for every model (Vals AI, Zapier, Proximal, Surge and CWE-bench). Argon leads 7 of those 9 and ties 1. Google ran 5 rows itself for all four models, and Argon leads 4. The other 5 pair a Google-run Argon score with rival numbers from elsewhere, and that’s where Argon loses most: 3 of its 5 losses sit in those mixed rows. If self-computing flattered Argon, you’d expect the opposite.
Where Argon trails, and why it matters for agentic coding
- FrontierSWE v2: 55.0% against Astra’s 65.5%. Last of four, on a leaderboard that scored every model.
- Terminal-bench 4.0: 57.4% against Opus 5.5’s 66.4%. Last again, though Astra and Fable 5.1 are within a point.
- PostTrainBench: 45.3% against Opus 5.5’s 49.3%. Second place, on a row Google ran for all models.
- Terminal-Bench Science 0.1: 57.6% against Astra’s 68.1%. Third, with a 6x verifier timeout on Argon’s run.
- OSWorld-2.0 (offline subset): 69.2% against Astra’s 72.6%. Anthropic reports only a combined score, so no Claude row.
Agentic coding splits 2 to 2. Argon wins DeepSWE v1.1 and Vibe Code Bench, then finishes last on FrontierSWE v2 and Terminal-bench 4.0. The DeepSWE win mixes harnesses: Argon ran on a mini-swe agent harness, Astra’s number is from the public leaderboard, and the Claude numbers are from system cards. Vibe Code Bench is cleaner, since Vals AI scored all four, but the margin is 1.6 points.
Cursor’s CursorBench result (52.8%, on par with Opus 5.5) was measured at default effort, not the maximum settings Google’s table compares.
For terminal-heavy agents or long repository tasks, weigh those two last-place rows above the headline count. Our GPT-6 Astra hands-on shows the FrontierSWE leader in use, and the Claude Fable 5.1 benchmarks breakdown covers Anthropic’s own launch numbers.
In launch week, Bloomberg reported that some Google employees said Argon underwhelms on coding and front-end work relative to its scores. Google called that inaccurate.
Methodology caveats that change how you read the table
- Max effort on both sides. Argon ran “with the Gemini API with the highest thinking settings.” For Astra, Fable 5.1 and Opus 5.5, Google defaults to the “maximum thinking/reasoning settings available.” That compares ceilings, not default behavior or cost. Argon’s API default is
mediumthinking, so default-setting results can differ. - LVBench frames. Gemini sampled video at 1 FPS; Astra got 800 frames, Fable 5.1 300 and Opus 5.5 600, “due to API limitations.” The models didn’t see identical inputs.
- OSWorld best of three. Argon’s score is “maxed over 3 runs with a single attempt per run,” a best run rather than an average.
- Safety filters on. Agent’s Last Exam (ALE-Claw harness, 5-hour window) and OSWorld ran with safety filters enabled, and flagged responses came back as empty strings. If anything, that works against Argon.
The cyber rows from DeepMind’s cyber page
The DeepMind cyber page adds four scores:
| Cyber eval | Gemini 4 Argon | Comparison |
|---|---|---|
| Real-world vulnerability discovery | 85.8% | Gemini 3.8 Flash Cyber 71.0% |
| Wiz Penetration Test Benchmark | 70.9% | Gemini 3.8 Flash Cyber 58.2% |
| Gray Swan IPI attack success (k=15, lower is better) | 0.7% | Lowest on the chart; Kimi K3 highest at 52.7% |
| CWE-bench v1 | 68% | Three-way tie with Grok 4.7 and GPT-6 Astra; Opus 5.5 at 67% |
The vulnerability eval used an internal Antigravity harness “that is not cyber specialized.” Both headline comparisons are against Google’s previous cyber model, not rivals. Our Argon cyber defense explainer covers what these mean for the APIs you run.
Third-party scoreboards
Artificial Analysis, Vals AI and Arena posted scores within minutes of the September 30 announcement, so they had pre-release access. Cursor and Scale AI published after general availability.
- Artificial Analysis, as of October 15, 2026, lists Gemini 4 Argon (High) at an Intelligence Index of 53, #8 of 216, at $1.99 per index task. Launch-week posts, ours included, quoted 52.56 and #8 of 223. The rank counts each reasoning setting separately; at launch, by distinct model, Argon tied Fable 5.1 and GPT-6 Astra behind Opus 5.5 (58 at max) and Sonnet 5.5 (56 at max), with a 15% hallucination rate (Astra at max, 51%) and about 62K output tokens per task against 27K for Astra at max.
- Vals AI put Argon at 68.90% on the Vals Index at launch, #1 of 41, at $15.68 per test, and listed a 262K max output for the configuration it tested. Google’s model page now sets 65,536 output tokens as the synchronous default, up to 1,048,576 (minus input) with background long decode.
- Arena ranked Argon #1 on Text at launch, at 1525, marked Preliminary on 4,942 votes, and #8 on WebDev.
- Cursor added Argon on October 13 and reports 52.8% on CursorBench at default effort, on par with Opus 5.5.
- Scale AI ranks Argon first on Humanity’s Sixth Sense (HSS) at 55.4%, ahead of GPT-6 Astra at 53.6%.
Why outlets publish 12, 13 and 14 wins
Recounting Google’s 19 rows: against the best of all three rivals, Argon wins 13, ties 1 and loses 5. Head to head, it’s 14-1-4 against Astra, 14-0-4 against Opus 5.5 (1 not reported) and 15-0-2 against Fable 5.1 (2 not reported). A 12 matches none of these framings, so check which rows that source counted.
How to rerun the rows you care about
Benchmarks describe Google’s harness; your prompts describe your product. With the API public, anyone with a paid key (there’s no free API tier) can reproduce a row or test their own workload.
- Pick real prompts that match the rows you care about: a repository fix, a terminal task, a long-document question.
- In Apidog, create an environment with
GEMINI_API_KEYandGEMINI_MODELset togemini-4-argon, plus a second environment for the model you run now. - Save each prompt as an Interactions request that reads
{{GEMINI_MODEL}}, grouped into a test scenario. - Assert on the output (status, required fields, expected content) and on a token ceiling sized to your cost cap, then keep the report as your baseline.
Each request is Google’s documented Interactions call, with the model in a variable:
MODEL="${GEMINI_MODEL:-gemini-4-argon}"
curl -s -X POST "https://generativelanguage.googleapis.com/v1beta/interactions" \
-H "x-goog-api-key: $GEMINI_API_KEY" \
-H "Content-Type: application/json" \
-d "{
\"model\": \"${MODEL}\",
\"input\": \"Your eval prompt here\",
\"generation_config\": {\"thinking_level\": \"high\"}
}"
Use high to match Google’s table, or the default medium to see what production traffic gets. Google’s migration notes matter here: temperature, top_p and top_k are ignored, so remove them and run each prompt more than once, and set client timeouts to at least 15 minutes, since hard prompts can take minutes.
At $2/$10 per 1M tokens through November 30, 2026, then $4/$20 from December 1, a request with 20,000 input and 4,000 output tokens (thinking included) costs $0.08, then $0.16. Tier 1 projects cap spend at $10 per rolling 10-minute window, so pace large batches. Our Argon vs Gemini 3.8 Flash comparison covers when the cheaper model is enough.
FAQ
Are Gemini 4 Argon’s benchmarks independently verified? Partly. Nine of the 19 rows come from public leaderboards for every model, and Artificial Analysis, Vals AI, Arena, Cursor and Scale AI have published their own results. The rest are Google-run for Argon.
Does Argon beat GPT-6 Astra on benchmarks? Head to head in Google’s table, Argon wins 14 rows, ties 1 and loses 4. On Scale AI’s HSS it leads 55.4% to 53.6%.
What is Gemini 4 Argon’s DeepSWE score? 77.9% on DeepSWE v1.1, ahead of Opus 5.5 (74.2%) and Astra (74.1%), on a Google-run harness.
Is Argon as good as Claude Opus 5.5 for coding? Mixed. Cursor calls its 52.8% CursorBench score on par with Opus 5.5, but Opus 5.5 leads Terminal-bench 4.0 by 9 points in Google’s own table.
Can I rerun these benchmarks myself? Yes, since October 13, 2026. Call gemini-4-argon on the paid Gemini API tier at high thinking; harness differences still apply. Our Argon release date guide covers the rollout.
Next step
Build the scenario, run it on gemini-4-argon at the current $2/$10 rate, and keep the report. You’ll have your own numbers before the December 1 price change. Download Apidog to set it up.



