Gemini 4 Carbon
| Released | 2026-11-10 |
|---|---|
| Price per 1M | $2.00 in · $10.00 out |
| Overall score | 67.2 · rank 2 |
| Context window | 1,048,576 tokens |
| Provider | |
| Weights | Proprietary |
Scores
| Dimension | Score | Rank | Benchmarks |
|---|---|---|---|
| Overall: Balanced | 67.2 | 2 | 34 |
| Overall: Coding | 67.3 | 1 | 28 |
| Overall: Reasoning | 68.4 | 2 | 16 |
| Overall: Agentic work | 66.7 | 1 | 25 |
| Overall: Frontend | 65.7 | 3 | 7 |
| Coding | 68.2 | 1 | 4 |
| Agents | 65.2 | 2 | 10 |
| Reasoning | 70.6 | 1 | 11 |
| Math | 66.6 | 5 | 3 |
| Knowledge | 66.9 | 2 | 2 |
| Frontend | 64.5 | 6 | 3 |
| General | 66.3 | 4 | 1 |
| Value vs price | +15.3 | 6 |
Buzz
| Posts on X | About 2,396 a day |
|---|---|
| Mood | 37% positive, 16% negative |
| Hacker News | 12 mentions this week |
| Checked | 2026-11-11 |
- Which model has the best overall intelligence? - Gemini 4 Carbon - Fable 5.2 - Opus 5.5 - GPT-6.2 Sol - GPT-6 Astra
- @Google released Gemini 4 Carbon plus a /swarm mode: a coordinator splits the task, workers edit code in parallel in separate git worktrees, verifiers run tests and challenge diffs. Try git worktree add today, no AI needed. These are Google's own claims. #AI #Gemini
- THE 10 BEST CODING IDES IN THE AGE OF CODEX AND CLAUDE CODE A coding IDE is no longer only a place you type. OpenAI Codex and Anthropic's Claude Code run as terminal agents that read a repo, edit files, run tests and open pull requests. The editor's job is to host those agents…
Configurations
| Configuration | Balanced | Benchmarks |
|---|---|---|
| Unspecified setting | 70.2 ±3.0 | 10 |
| High | 66.4 ±1.1 | 24 |
| Low | 65.4 ±4.4 | 4 |
| Medium | 68.4 ±4.2 | 4 |
Evidence
Coding 68.2 ±2.9
| Benchmark | Result | Source | Date |
|---|---|---|---|
| FrontierSWE | 61.2% | Epoch AI | 2026-11-11 |
| CursorBench | 60.4% | Epoch AI | 2026-11-11 |
| LiveBench 2.0 Coding | 91.0% | LiveBench | 2026-11-06 |
| SciCode | 64.2% | Epoch AI | 2026-11-11 |
Agents 65.2 ±1.7
| Benchmark | Result | Source | Date |
|---|---|---|---|
| Terminal-Bench 4.0 (Vals run) | 63.1% | Vals AI | 2026-11-09 |
| APEX-Agents | 84.2% | Epoch AI | 2026-11-11 |
| LiveBench 2.0 Agentic | 72.7% | LiveBench | 2026-11-06 |
| LiveBench 2.0 Browser Use | 80.0% | LiveBench | 2026-11-06 |
| EMB v1.1 (Excel financial modeling) | 76.5% | Vals AI | 2026-11-09 |
| LiveBench 2.0 Research | 92.5% | LiveBench | 2026-11-06 |
| Tax Agent Bench | 51.8% | Vals AI | 2026-11-09 |
| Finance Agent v2.1 | 67.2% | Vals AI | 2026-11-09 |
| Vending-Bench 2 | $16,484 | Epoch AI | 2026-11-11 |
| HLAB-Vals 1.1 (legal agent) | 17.5% | Vals AI | 2026-11-09 |
Reasoning 70.6 ±1.8
| Benchmark | Result | Source | Date |
|---|---|---|---|
| ARC-AGI-2 | 94.2% | Epoch AI | 2026-11-11 |
| ARC-AGI-2 (ARC Prize verified) | 94.2% | ARC Prize | 2026-11-11 |
| Extended NYT Connections | 99.3% | Lech Mazur | 2026-11-10 |
| GPQA Diamond | 96.5% | Epoch AI | 2026-11-02 |
| CritPt | 36.0% | Epoch AI | 2026-11-11 |
| ARC-AGI-1 (ARC Prize verified) | 98.5% | ARC Prize | 2026-11-11 |
| ARC-AGI-1 | 98.5% | Epoch AI | 2026-11-11 |
| SimpleBench | 92.6% | SimpleBench | 2026-11-10 |
| ARC-AGI-3 | 99.1% | ARC Prize | 2026-11-11 |
| Chess Puzzles | 78.0% | Epoch AI | 2026-11-02 |
| LiveBench 2.0 Reasoning | 90.6% | LiveBench | 2026-11-06 |
| Mystery Game Puzzles | 73.0% | Epoch AI | 2026-11-02 |
| LiveBench 2.0 Data Analysis | 86.5% | LiveBench | 2026-11-06 |
Math 66.6 ±2.0
| Benchmark | Result | Source | Date |
|---|---|---|---|
| FrontierMath Tiers 1-3 | 90.5% | Epoch AI | 2026-11-02 |
| FrontierMath Tier 4 | 95.1% | Epoch AI | 2026-11-02 |
| OTIS Mock AIME 2024-2025 | 100.0% | Epoch AI | 2026-11-02 |
Knowledge 66.9 ±2.7
| Benchmark | Result | Source | Date |
|---|---|---|---|
| SimpleQA Verified | 77.7% | Epoch AI | 2026-11-02 |
| Legal Research Bench | 55.8% | Vals AI | 2026-11-09 |
Frontend 64.5 ±2.8
| Benchmark | Result | Source | Date |
|---|---|---|---|
| LiveBench 2.0 Design | 80.2% | LiveBench | 2026-11-06 |
| Vibe Code Bench v1.1 | 89.1% | Vals AI | 2026-11-09 |
| LMArena WebDev | 1,768 | LMArena | 2026-11-09 |
General 66.3 ±4.1
| Benchmark | Result | Source | Date |
|---|---|---|---|
| LMArena Text | 1,526 | LMArena | 2026-11-09 |
Results not used in the score
| Benchmark | Result | Source | Why not used |
|---|---|---|---|
| ARC-AGI-1 | 97.5% | Epoch AI | A stronger configuration result was used |
| ARC-AGI-1 | 85.5% | Epoch AI | A stronger configuration result was used |
| ARC-AGI-1 (ARC Prize verified) | 97.5% | ARC Prize | A stronger configuration result was used |
| ARC-AGI-1 (ARC Prize verified) | 85.5% | ARC Prize | A stronger configuration result was used |
| ARC-AGI-2 | 92.1% | Epoch AI | A stronger configuration result was used |
| ARC-AGI-2 | 49.6% | Epoch AI | A stronger configuration result was used |
| ARC-AGI-2 (ARC Prize verified) | 92.1% | ARC Prize | A stronger configuration result was used |
| ARC-AGI-2 (ARC Prize verified) | 49.6% | ARC Prize | A stronger configuration result was used |
| ARC-AGI-3 | 96.9% | ARC Prize | A stronger configuration result was used |
| ARC-AGI-3 | 63.9% | ARC Prize | A stronger configuration result was used |
| ARC-AGI-3 | 65.1% | ARC Prize | A stronger configuration result was used |
| ARC-AGI-3 | 97.7% | ARC Prize | A stronger configuration result was used |
| CursorBench | 60.1% | Epoch AI | A stronger configuration result was used |
| CursorBench | 58.6% | Epoch AI | A stronger configuration result was used |
Names matched to this model
gemini-4-carbon