GoogleGemini 4 Carbon
As of 29 Oct 2026, Gemini 4 Carbon by Google scores 66.4 on the BenchLeader Index (top 25 of 587 ranked configurations), with its best results in reasoning (79) and weakest in long context (65). It costs $8.00 per 1M tokens blended ($4.00 in / $20.00 out), and streams about 100 tokens/s (Artificial Analysis). Released 28 Oct 2026; 7 independent results so far.
Key facts (as of 29 Oct 2026)
- Gemini 4 Carbon costs $8.00 per 1M tokens blended ($4.00 in / $20.00 out) (list price, 29 Oct 2026).
- Gemini 4 Carbon streams about 100 tokens/s (Artificial Analysis, 29 Oct 2026).
- Gemini 4 Carbon was released on 28 Oct 2026 (Google).
In detail. Gemini 4 Carbon is a Google proprietary reasoning model, released 28 Oct 2026. Its best configuration (high reasoning effort) ranks #11 of 587 on the BenchLeader Index at 66.4 ±5.2, in the top 25. The ± is the point: 69 other configurations score within that range, so they and this one cannot be told apart on quality alone — price and speed are what separate them. It scores highest in reasoning (79) and lowest in long context (65). At $8.00 per million tokens blended it is among the most expensive ranked models. Output speed of 100 tokens per second puts it faster than most, with a first answer in 103.1 s. It has been measured at 2 reasoning-effort settings; this summary describes the best-scoring one, and the tabs above switch between them. Last measured 28 Oct 2026.
- Blended price
- $8.00/M
- $4.00 in · $20.00 out
- Output speed
- 100 tok/s
- measured by Artificial Analysis
- First answer
- 103 s
- Context
- 1M
- Full answer
- 108 s
- median, reasoning included
- Cost per run
- $7.76
- one full Intelligence Index run
- Released
- 28 Oct 2026
- Overall index66
- Reasoning79
- Coding72
- Knowledge76
- Long context65
- Composite76
Reasoning-effort configurations
The same model behaves differently depending on how much it is allowed to think. Each row is one setting, scored only on the benchmarks that were run at that setting. “Not stated” collects results from publishers that did not say which setting they used; for a reasoning model that is usually its thinking mode, but we do not assume it. Pick a setting here or at the top of the page to see its price, speed and category scores.
| Effort | Index | Rank | Speed | First answer | Chat reply cost | Categories |
|---|---|---|---|---|---|---|
| highbest | 66.4 | #11 | 100 tok/s | 103 s | $0.0076 | Coding 72 · Composite 76 · Knowledge 76 · Long context 65 · Reasoning 79 |
| not stated | – | – | – | – | – |
Benchmark results
See all LLM benchmarks. One column per reasoning effort. Rank is among every configuration of every model on that benchmark. Hover a score for the run it came from.
Reasoning
| Benchmark | high | Source |
|---|---|---|
| Humanity's Last Exam (AA)not in index | 60.4%#2 | Artificial Analysis |
| HLE-Diamond (AA) | 69.1%#1 | Artificial Analysis |
| CritPtnot in index | 36.0%#1 | Artificial Analysis |
| CritPt v2 hard subset (AA) | 63.0%#10 | Artificial Analysis |
Coding
| Benchmark | high | Source |
|---|---|---|
| SciCode | 64.2%#3 | SciCode |
| SciCode (AA)not in index | 64.2%#3 | Artificial Analysis |
| Vibe Code Bench v1.1not in index | 89.1%#8 | Vals AI |
| Terminal-Bench 4.0 (AA)not in index | 65.2%#2 | Artificial Analysis |
| FrontierCode 1.1 (AA)not in index | 48.5%#7 | Artificial Analysis |
| FrontierCode 2.0 (AA) | 40.6%#8 | Artificial Analysis |
Agents & tools
| Benchmark | high | not stated | Source |
|---|---|---|---|
| Finance Agent v2.1not in index | 67.2%#5 | – | Vals AI |
| Harvey's Legal Agent Benchmark v1.1not in index | 17.5%#6 | – | Vals AI |
| Legal Research Benchnot in index | 55.8%#1 | – | Vals AI |
| Tax Agent Benchnot in index | 48.9%#3 | – | Vals AI |
| Terminal-Bench 4.0 (Vals)not in index | 63.1%#4 | – | Vals AI |
| Vending-Bench 2not in indexeffort not stated | – | 16484.2#1 | Andon Labs |
| AutomationBenchnot in index | 77.8%#1 | – | Artificial Analysis |
| GDP.pdf (AA)not in index | 28.4%#15 | – | Artificial Analysis |
| AA-Briefcase v1.1not in index | 1609#9 | – | Artificial Analysis |
| Terminal-Bench-Science 0.1 (AA)not in index | 64.8%#1 | – | Artificial Analysis |
Knowledge
| Benchmark | high | Source |
|---|---|---|
| AA-Omniscience v1.1 | 53.5#1 | Artificial Analysis |
| Excel Modeling Benchmark v1.1not in index | 76.5%#2 | Vals AI |
| AA-Omniscience v1.1: accuracynot in index | 58.3%#21 | Artificial Analysis |
| AA-Omniscience v1.1: non-hallucinationnot in index | 88.6%#2 | Artificial Analysis |
Multimodal
| Benchmark | not stated | Source |
|---|---|---|
| Blueprint-Bench 2not in index | 58.6%#1 | Epoch AI Benchmarking Hub |
Long context
| Benchmark | high | Source |
|---|---|---|
| AA-LCR v1.1 | 79.3%#103 | Artificial Analysis |
Composite
| Benchmark | high | Source |
|---|---|---|
| AA Intelligence Index v5.0 | 59.4#1 | Artificial Analysis |
| AA Progress Index v1not in index | 351.6#1 | Artificial Analysis |
| Vals Indexnot in index | 70.4#2 | Vals AI |
| Vals Finance Indexnot in index | 66.3%#1 | Vals AI |
Where it wins
Benchmarks where this configuration ranks in the top five of every configuration measured.
What a task costs
Estimates from list price, output speed and time to first answer for the best configuration. “With caching” assumes three-quarters of the input is served from the prompt cache. Reasoning tokens are not modelled.
| Workload | Tokens in / out | Cost | With caching | Time |
|---|---|---|---|---|
| Chat reply | 400 / 300 | $0.0076 | – | 1.8 min |
| Summarize a 30-page report | 12,000 / 600 | $0.060 | – | 1.8 min |
| Code edit | 6,000 / 1,500 | $0.054 | – | 2.0 min |
| Agentic coding session | 60,000 / 4,000 | $0.320 | – | 2.4 min |
| Structured extraction | 2,000 / 200 | $0.012 | – | 1.8 min |
More from Google
Compare Gemini 4 Carbon with…
- Gemini 4 Carbon vs Claude Opus 5.5
- Gemini 4 Carbon vs GPT-6 Astra
- Gemini 4 Carbon vs GPT-6.1 Sol
- Gemini 4 Carbon vs Claude Fable 5.1
- Gemini 4 Carbon vs Claude Fable 5
- Gemini 4 Carbon vs Gemini 4 Argon
- Gemini 4 Carbon vs Claude Opus 5
- Gemini 4 Carbon vs Claude Sonnet 5.5
- Gemini 4 Carbon vs GPT-5.6 Sol
- Gemini 4 Carbon vs GPT-6.1 Terra
- Gemini 4 Carbon vs GPT-5.5
- Gemini 4 Carbon vs Muse Spark
- Gemini 4 Carbon vs Muse Wonder
- Gemini 4 Carbon vs GPT-6 Sol
- Gemini 4 Carbon vs GPT-5.4
- Gemini 4 Carbon vs Claude Opus 4.7
- Gemini 4 Carbon vs GPT-5.4 Pro
- Gemini 4 Carbon vs GPT-5.3-Codex
- Gemini 4 Carbon vs Claude Opus 4.6
- Gemini 4 Carbon vs Gemini 3.5 Flash
- Gemini 4 Carbon vs Muse Wonder Full
- Gemini 4 Carbon vs Qwen3.6 Max
- Gemini 4 Carbon vs Gemini 3.7 Flash
- Gemini 4 Carbon vs Muse Spark 1.1
- Gemini 4 Carbon vs Kimi K3
- Gemini 4 Carbon vs Kimi K3.1 Pro
- Gemini 4 Carbon vs Muse Spark 1.2
- Gemini 4 Carbon vs Gemini 3.6 Flash
- Gemini 4 Carbon vs GPT-5.6 Terra
- Gemini 4 Carbon vs o3
- Gemini 4 Carbon vs Muse Spark 1.3
- Gemini 4 Carbon vs Claude Opus 4.8
- Gemini 4 Carbon vs Gemini 3.1 Pro
- Gemini 4 Carbon vs Grok 4.6
- Gemini 4 Carbon vs Qwen3.7
- Gemini 4 Carbon vs Gemini 3 Pro
- Gemini 4 Carbon vs Gemini 3.8 Flash
- Gemini 4 Carbon vs GLM 5.2
- Gemini 4 Carbon vs GLM-5
- Gemini 4 Carbon vs GPT-5.2
- Gemini 4 Carbon vs Grok 4.5
- Gemini 4 Carbon vs GPT-5 Pro
- Gemini 4 Carbon vs Kimi K2.6
- Gemini 4 Carbon vs Claude Opus 4.5
- Gemini 4 Carbon vs Claude Sonnet 5
- Gemini 4 Carbon vs Mimo v2 Pro
- Gemini 4 Carbon vs Qwen3.6 Plus
- Gemini 4 Carbon vs Qwen3.8 Max
- Gemini 4 Carbon vs Gemini 3 Flash
- Gemini 4 Carbon vs Mimo v2 Omni
- Gemini 4 Carbon vs Kimi K2.5
- Gemini 4 Carbon vs GPT-5-Codex
- Gemini 4 Carbon vs Grok 4.20
- Gemini 4 Carbon vs GPT-5.6 Luna
- Gemini 4 Carbon vs Muse Spark Open 816B A42B
- Gemini 4 Carbon vs Dola Seed 2.0 Pro
- Gemini 4 Carbon vs Grok 4.20 Multi-Agent
- Gemini 4 Carbon vs Grok 4
- Gemini 4 Carbon vs Claude Sonnet 4.6
See also
Questions people ask
- What are Gemini 4 Carbon's benchmark scores?
- As of 29 Oct 2026, Gemini 4 Carbon (high effort) scores 66.4 ±5.2 on the BenchLeader Index, top 25, strongest in reasoning (79) and weakest in long context (65). That rests on 7 independent benchmark results, and every score on this page links to its publisher.
- How does Gemini 4 Carbon rank?
- Gemini 4 Carbon (high effort) is #11 of 587 ranked configurations on 29 Oct 2026 (top 25). 69 other configurations score within its margin, so the exact position matters less than the band. The leader is Claude Opus 5.5 (max effort) at 69.6.
- How much does Gemini 4 Carbon cost per 1M tokens?
- As of 29 Oct 2026, Gemini 4 Carbon costs $4.00 per 1M input tokens and $20.00 per 1M output tokens, $8.00 blended (three input tokens per output token).
- Is Gemini 4 Carbon better than Claude Fable 5?
- As of 29 Oct 2026, Claude Fable 5 and Gemini 4 Carbon (high) are neck and neck on quality (66.6 vs 66.4 on the BenchLeader Index, inside the ±5.9 margin of error), and Gemini 4 Carbon (high) costs 2.5× less, which makes it the better buy for most workloads. The head-to-head page has the benchmark-by-benchmark detail.
- When was Gemini 4 Carbon released?
- Google released Gemini 4 Carbon on 28 Oct 2026. BenchLeader has results for it at 2 reasoning-effort settings.
Data as of 29 Oct 2026. Compare these configurations.
Cite as: BenchLeader, “Gemini 4 Carbon: benchmarks, pricing, speed and rank”, data as of 29 Oct 2026, https://www.benchleader.com/models/gemini-4-carbon.