BlogLLM Benchmark Scores as Published, Side by Side
Gemini 4 Carbon: Benchmark Scores as Published
Every result the catalog holds for Gemini 4 Carbon: 21 tests in 11 areas, from 3 pages. All were printed by Google itself. Each number is as printed; nothing is ranked or rescored.
Developer GoogleReleased 2026-10-28Data version 2026103001, as of 2026-10-30
Knowledge Work
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| AutomationBench | 53.9% | Google, Gemini 4 Carbonchecked 2026-10-29 | - |
| Vending-Bench 2 (final balance in USD) | $16484 | Google, Gemini 4 Carbonchecked 2026-10-29 | - |
| Vals Index v2.2 | 70.4% | Google, Gemini 4 Carbonchecked 2026-10-29 | - |
Agentic Coding
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| Terminal-Bench 4.0 | 65.2% | Google, Gemini 4 Carbonchecked 2026-10-29 | - |
| FrontierSWE v2 | 61.2% | Google, Gemini 4 Carbonchecked 2026-10-29 | - |
| Code Arena | 1768 | Google, Gemini 4 Carbon: our next frontier modelchecked 2026-10-29 | - |
| SWE-sweep | 9.6% | Google, Gemini 4 Carbonchecked 2026-10-29 | Harness: mini-swe-agent |
Science and Math
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| Terminal-Bench Science 0.1 | 64.8% | Google, Gemini 4 Carbonchecked 2026-10-29 | - |
| RiemannBench | 80% | Google, Gemini 4 Carbonchecked 2026-10-29 | - |
Long Context
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| GraphWalks BFS 256k to 1M (F1) | 96.2% | Google, Gemini 4 Carbonchecked 2026-10-29 | - |
Computer Use
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| OSWorld 2.1 offline subset | 81.5% | Google, Gemini 4 Carbonchecked 2026-10-29 | Harness: Gemini CUA harness |
Multimodal Understanding
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| Chartography (no tools) | 72.5% | Google, Gemini 4 Carbonchecked 2026-10-29 | - |
Cybersecurity
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| ExploitGym | 54.8% | Google, Gemini 4 Carbonchecked 2026-10-29; printed on 1 more page | Harness: Antigravity |
| SEC-Bench Pro | 88.2% | Google, Gemini 4 Carbonchecked 2026-10-29; printed on 1 more page | Harness: Antigravity |
| CWE-bench v1 | 71% | Google, Gemini 4 Carbonchecked 2026-10-29; printed on 1 more page | - |
Tool Use and Workflows
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| Agents' Last Exam v1.1 (score) | 77.4% | Google, Gemini 4 Carbonchecked 2026-10-29 | Harness: ALE-Claw |
Safety and Alignment
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| Gray Swan IPI (attack success rate after k attempts)attack success rate after k attempts, lower is better | 0% | Google, Gemini 4 Carbon: our next frontier modelchecked 2026-10-29 | - |
| Terminal-Bench 4.0 cheating attempt ratelower is better | 0.4% | Google, Gemini 4 Carbon: our next frontier modelchecked 2026-10-29 | - |
| Terminal-Bench 4.0 successful cheating ratelower is better | 0% | Google, Gemini 4 Carbon: our next frontier modelchecked 2026-10-29 | - |
Engineering and Design
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| Blueprint-Bench 2 (normalized score) | 58.6% | Google, Gemini 4 Carbonchecked 2026-10-29 | - |
Preference and Composite Rankings
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| Artificial Analysis Intelligence Index v5.0 | 59 | Google, Gemini 4 Carbonchecked 2026-10-29; printed on 1 more page | - |
Price
API list prices in US dollars per one million tokens, as printed on the developer's pricing page on the day checked.
| Price | Value | Printed by | Setup as printed |
|---|---|---|---|
| Input | $2 | Google, Gemini 4 Carbon launch pricingchecked 2026-10-29 | Offer: promotional price |
| Output | $10 | Google, Gemini 4 Carbon launch pricingchecked 2026-10-29 | Offer: promotional price |
Side by side
More from Google
Every number here is read from the same catalog file as the benchmark table, data version 2026103001. Where one page printed the model under several setups, each value is listed. The catalog is public in armanobosyan/sugra-llm-benchmarks; a wrong number can be reported from the table, with Report an error under the value, or as an issue in that repository.