Gemini 3.8 Flash vs Gemini 4 Carbon
Full profiles: Gemini 3.8 Flash (Google) and Gemini 4 Carbon (Google): every benchmark, price and rank.
Updated · Data from 10 sources
As of 29 Oct 2026, Gemini 3.8 Flash and Gemini 4 Carbon are neck and neck on quality (61.8 vs 65.1 on the BenchLeader Index, inside the ±7.1 margin), and Gemini 3.8 Flash is 5.3× cheaper at $1.50 vs $8.00 per 1M tokens blended. Choose 3.8 Flash for long context (66 vs 65) and fast first answers (26.7 s vs 103.1 s), where it also streams faster; choose 4 Carbon for reasoning (79 vs 64), coding (72 vs 60), and knowledge (76 vs 64). Gemini 3.8 Flash's current price runs until 31 Dec 2026, then $3.00.
Key facts (as of 29 Oct 2026)
- Gemini 3.8 Flash scores 61.8 on the BenchLeader Index and Gemini 4 Carbon scores 65.1 (BenchLeader, 29 Oct 2026).
- Gemini 3.8 Flash costs $1.50 per 1M tokens blended until 31 Dec 2026, then $3.00 (https://ai.google.dev/gemini-api/docs/pricing, checked 20 Oct 2026).
- Gemini 4 Carbon costs $8.00 per 1M tokens blended (list price, 29 Oct 2026).
- Gemini 3.8 Flash streams faster than Gemini 4 Carbon (Artificial Analysis, 7-day median, 29 Oct 2026).
- Of the 25 benchmarks both models have, Gemini 4 Carbon leads 24 (BenchLeader, 29 Oct 2026).
Gemini 3.8 Flash benchmarks high Google Gemini 4 Carbon benchmarks high GoogleOpen in the comparison toolWhich to pick for…
The pick's numbers come first. Category scores on the BenchLeader scale (50 is the average model, 15 points one standard deviation). High confidence means a lead of half a standard deviation or more; under one point is called either.
| Use case | Pick | Why | Confidence |
|---|---|---|---|
| Coding | Gemini 4 Carbon | 72 vs 60 category score; FrontierCode 2.0 (AA) 40.6% vs 18.7% | High |
| Reasoning | Gemini 4 Carbon | 79 vs 64 category score; HLE-Diamond (AA) 69.1% vs 33.1% | High |
| Knowledge | Gemini 4 Carbon | 76 vs 64 category score; AA-Omniscience v1.1 53.5 vs 23.3 | High |
| Long documents | Gemini 3.8 Flash | 66 vs 65 category score; AA-LCR v1.1 81.3% vs 79.3% | Medium |
| Chat quality (human preference) | – | Only Gemini 3.8 Flash has an Arena rating (69 category score) | Not comparable |
| Lowest cost | Gemini 3.8 Flash | $1.50 vs $8.00 per 1M blended; agentic coding session $0.060 vs $0.320; from 1 Jan 2027: $3.00 vs $8.00 | – |
| Fastest first answer | Gemini 3.8 Flash | 26.7 s (Gemini 3.8 Flash (high)) vs 103.1 s (Gemini 4 Carbon (high)) to the first answer token | – |
| High volume on a budget | Neither flagship | A smaller model is usually the better fit: Cheapest good LLMs, Fastest LLMs | – |
Key differences
- Gemini 3.8 Flash (high) and Gemini 4 Carbon (high) are neck and neck on quality (61.8 vs 65.1, within the ±7.1 margin).
- Gemini 3.8 Flash (high) is stronger in agents & tools, human preference, long context, maths, multimodal.
- Gemini 4 Carbon (high) is stronger in coding, knowledge, reasoning.
- Gemini 3.8 Flash (high) is 5.3× cheaper ($1.50 vs $8.00 per 1M blended) until 31 Dec 2026, then $3.00 vs $8.00 from 1 Jan 2027.
- Gemini 3.8 Flash (high) streams 1.4× faster (143 vs 100 tokens per second).
At a glance
| Spec | Gemini 3.8 Flash | Gemini 4 Carbon |
|---|---|---|
| Developer | ||
| Released | 2 Sept 2026 | 28 Oct 2026 |
| Context window (tokens) | 1,048,576 | 1,000,000 |
| Compared at | Gemini 3.8 Flash (high) | Gemini 4 Carbon (high) |
| Open weights | No | No |
| Price in / out / cached input per 1M | $0.750 / $3.75 / $0.075 | $4.00 / $20.00 / – |
| Announced price change | $1.50 / $7.50 from 1 Jan 2027 | None announced |
Limits, availability and announced prices from the vendors' own pages: ai.google.dev pricing (checked 20 Oct 2026). Everything else is from the day's data.
Category by category
Category scores on the BenchLeader scale: 50 is the average evaluated model, 15 points is one standard deviation. A gap under one point is called level.
| Category | Gemini 3.8 Flash (high) | Gemini 4 Carbon (high) | Leader |
|---|---|---|---|
| Reasoning | 63.5 | 78.9 | Gemini 4 Carbon |
| Coding | 59.6 | 71.7 | Gemini 4 Carbon |
| Agents & tools | 51.4 | – | not comparable |
| Maths | 62.3 | – | not comparable |
| Knowledge | 64.2 | 75.8 | Gemini 4 Carbon |
| Human preference | 69.0 | – | not comparable |
| Multimodal | 67.5 | – | not comparable |
| Long context | 65.5 | 64.5 | Gemini 3.8 Flash |
Where sources disagree
Across the 25 benchmarks both have, Gemini 3.8 Flash leads 1, Gemini 4 Carbon leads 24. Every overall index that rates both puts Gemini 4 Carbon ahead.
| Overall index | Gemini 3.8 Flash (high) | Gemini 4 Carbon (high) | Leader |
|---|---|---|---|
| BenchLeader Index | 61.8 | 65.1 | neck and neck |
| AA Intelligence Index v5.0 | 30.6 | 59.4 | Gemini 4 Carbon |
| Vals Index | 56.0 | 70.4 | Gemini 4 Carbon |
Each publisher runs its own tasks, harnesses and effort settings, so their overall verdicts can differ; BenchLeader reconciles them on one scale. † measured at another effort setting of the same model (the publisher did not test the one compared here).
Price, speed and latency
List prices blended at three input tokens per output token. Per-task prices use the token counts on the methodology page; suite cost is what Artificial Analysis paid to run its whole Intelligence Index.
| Measure | Gemini 3.8 Flash (high) | Gemini 4 Carbon (high) |
|---|---|---|
| Blended price per 1M tokens, until 31 Dec 2026 | $1.50 | $8.00 |
| Input / output per 1M | $0.750 / $3.75 | $4.00 / $20.00 |
| Blended price per 1M from 1 Jan 2027 | $3.00 | $8.00 |
| Input / output per 1M from 1 Jan 2027 | $1.50 / $7.50 | $4.00 / $20.00 |
| Cached input per 1M | $0.075 | – |
| Price per task: chat reply | $0.0014 | $0.0076 |
| Price per task: summarize a 30-page report | $0.011 | $0.060 |
| Price per task: code edit | $0.010 | $0.054 |
| Price per task: agentic coding session | $0.060 | $0.320 |
| Price per task: structured extraction | $0.0023 | $0.012 |
| Cost to run a full benchmark suite | $7.77 | $7.76 |
| Output speed | 143 tok/s | 100 tok/s |
| Time to first answer | 26.7 s | 103.1 s |
| Time to a complete answer | 30.2 s | 108.1 s |
Per token is not per task: on Artificial Analysis' full suite Gemini 3.8 Flash (high) cost $7.77 and Gemini 4 Carbon (high) $7.76, because Gemini 3.8 Flash (high) used about 5.3× as many tokens for the same work, which a per-token price hides.
Each model is compared at its best-ranked reasoning-effort setting. Gemini 3.8 Flash (high) scores 61.8 and takes 27 s to its first answer token; Gemini 3.8 Flash (medium) scores 60.3 in 3 s, Gemini 3.8 Flash (low) scores 55.6 in 3 s. Gemini 4 Carbon (high) scores 65.1 and takes 103 s to its first answer token. Higher effort means more thinking before the answer: better scores, longer waits and more tokens billed. How reasoning effort changes scores and cost.
Every benchmark
140 benchmarks, grouped by category; each name links to the board and its leaders. † measured at another effort setting of the same model, shown for reference and not counted as a win.
| Metric | Publisher | Gemini 3.8 Flash (high) | Gemini 4 Carbon (high) |
|---|---|---|---|
| BenchLeader Index | BenchLeader | 61.8 | 65.1 |
| Agents & tools score | BenchLeader | 51.4 | – |
| Coding score | BenchLeader | 59.6 | 71.7 |
| Human preference score | BenchLeader | 69.0 | – |
| Knowledge score | BenchLeader | 64.2 | 75.8 |
| Long context score | BenchLeader | 65.5 | 64.5 |
| Maths score | BenchLeader | 62.3 | – |
| Multimodal score | BenchLeader | 67.5 | – |
| Reasoning score | BenchLeader | 63.5 | 78.9 |
| Blended price $/M | BenchLeader | $1.50 | $8.00 |
| Output speed | BenchLeader | 143 tok/s | 100 tok/s |
| Time to first answer | BenchLeader | 26.7 s | 103.1 s |
| Context window | BenchLeader | 1,048,576 tokens | 1,000,000 tokens |
| Reasoning | |||
| GPQA Diamond | Epoch AI Benchmarking Hub | 95.4% | – |
| Humanity's Last Exam | Scale AI / CAIS | 44.5%† | – |
| Arena Hard Prompts | Arena (LMArena) | 1517 | – |
| LiveBench 2.0 Reasoning | LiveBench | 89.3%† | – |
| LiveBench 1.0 Reasoning | LiveBench | 89.3% | – |
| GPQA Diamond (AA) | Artificial Analysis | 95.3% | – |
| Humanity's Last Exam (AA) | Artificial Analysis | 47.8% | 60.4% |
| HLE-Diamond (AA) | Artificial Analysis | 33.1% | 69.1% |
| GPQA Diamond (Vals) | Vals AI | 94.4% | – |
| ARC-AGI-1 | ARC Prize | 98.5% | – |
| ARC-AGI-2 | ARC Prize | 89.2% | – |
| ARC-AGI-3 | ARC Prize | 35.0% | – |
| CritPt | Artificial Analysis | 18.3% | 36.0% |
| CritPt v2 hard subset (AA) | Artificial Analysis | 43.5% | 63.0% |
| MysteryMechanism | Vals AI | 36.5% | – |
| Chess Puzzles | Epoch AI Benchmarking Hub | 58.0% | – |
| Mystery Game Puzzles | Epoch AI Benchmarking Hub | 47.0% | – |
| LMCA | Epoch AI Benchmarking Hub | 51.5%† | – |
| DTBench | Epoch AI Benchmarking Hub | 95.7%† | – |
| Coding | |||
| SciCode | SciCode | 56.6% | 64.2% |
| WeirdML | WeirdML | 84.8% | – |
| FrontierCode | Cognition | 41.2%† | – |
| Arena Coding | Arena (LMArena) | 1528 | – |
| Arena WebDev | Arena (LMArena) | 1584 | – |
| LiveBench 2.0 Coding | LiveBench | 82.2%† | – |
Show all 140 benchmarks
| Metric | Publisher | Gemini 3.8 Flash (high) | Gemini 4 Carbon (high) |
|---|---|---|---|
| Coding | |||
| LiveBench 1.0 Coding | LiveBench | 72.5% | – |
| LiveCodeBench | Vals AI | 89.5% | – |
| IOI | Vals AI | 56.9% | – |
| SWE-bench (Vals) | Vals AI | 80.0% | – |
| Code Migration | Vals AI | 36.5% | – |
| Code Migration v1.1 | Vals AI | 38.6% | – |
| ProgramBench | Vals AI | 1.0% | – |
| SWE-bench Multimodal | Vals AI | 40.8% | – |
| Vibe Code Bench 1-100 | Vals AI | 18.8% | – |
| Vibe Code Bench v1.1 | Vals AI | 78.7% | 89.1% |
| SWE Atlas: Codebase QnA | Scale AI SEAL | 47.0%† | – |
| SWE Atlas: Refactoring | Scale AI SEAL | 44.8%† | – |
| SWE Atlas: Test Writing | Scale AI SEAL | 53.7%† | – |
| Surface Evolver Bench | Epoch AI Benchmarking Hub | 76.9% | – |
| DeepSWE v1.1 | Epoch AI Benchmarking Hub | 73.8% | – |
| CursorBench | Cursor | 39.6% | – |
| ALE-Bench | Epoch AI Benchmarking Hub | 1269.7 | – |
| FrontierSWE | Epoch AI Benchmarking Hub | 19.6% | – |
| Terminal-Bench 4.0 (AA) | Artificial Analysis | 19.7% | 65.2% |
| Terminal-Bench 2.1 (AA) | Artificial Analysis | 87.6% | – |
| Startrise Frontend | Startrise | 64.6%† | – |
| FrontierCode 1.1 (AA) | Artificial Analysis | 36.4% | 48.5% |
| FrontierCode 2.0 (AA) | Artificial Analysis | 18.7% | 40.6% |
| Agents & tools | |||
| Terminal-Bench | Terminal-Bench | 19.1% | – |
| Remote Labor Index | Scale AI / CAIS | 5.8%† | – |
| APEX-Agents | Mercor | 64.3%† | – |
| Arena Agent | Arena (LMArena) | 3.2 | – |
| LiveBench Long Agentic | LiveBench | 16.5%† | – |
| LiveBench Browser | LiveBench | 45.1%† | – |
| LiveBench Agentic Coding | LiveBench | 54.2% | – |
| Terminal-Bench 2.1 (Vals) | Vals AI | 81.3% | – |
| HiL-Bench | Scale AI SEAL | 41.5%† | – |
| GDPval-AA v2.1 | Artificial Analysis | 46.8% | – |
| τ³-Banking (AA) | Artificial Analysis | 45.0% | – |
| ITBench SRE (AA) | Artificial Analysis | 52.5% | – |
| Analyst Agent (AA) | Artificial Analysis | 46.3% | – |
| CUA-bench | Vals AI | 4.2% | – |
| CyberBench | Vals AI | 43.8% | – |
| Finance Agent v2 | Vals AI | 61.4% | – |
| Finance Agent v2.1 | Vals AI | 65.8% | 67.2% |
| Harvey's Legal Agent Benchmark v1.1 | Vals AI | 12.1% | 17.5% |
| Harvey's Legal Agent Benchmark v1 | Vals AI | 10.0% | – |
| Legal Research Bench | Vals AI | 38.9% | 55.8% |
| SkillsBench | Vals AI | 58.0% | – |
| SREBench | Vals AI | 11.1% | – |
| Tax Agent Bench | Vals AI | 32.1% | 48.9% |
| Terminal-Bench 4.0 (Vals) | Vals AI | 19.7% | 63.1% |
| Terminal-Bench Science | Vals AI | 8.6% | – |
| Time Horizon Index: KSP | Vals AI | 18.8% | – |
| Vending-Bench 2 | Andon Labs | 5093.8† | 16484.2† |
| AutomationBench | Artificial Analysis | 59.9% | 77.8% |
| GDP.pdf (AA) | Artificial Analysis | 21.0% | 28.4% |
| Harvey LAB-AA v1.1 | Harvey | 0.0% | – |
| AA-Briefcase v1.1 | Artificial Analysis | 1202 | 1609 |
| Terminal-Bench-Science 0.1 (AA) | Artificial Analysis | 10.0% | 64.8% |
| CWE-Bench (AA) | Artificial Analysis | 52.5% | – |
| DeepSec-Bench (AA) | Artificial Analysis | 14.5% | – |
| CyberGym end-to-end (AA) | Artificial Analysis | 8.4% | – |
| FACTS Search | 76.3%† | – | |
| Maths | |||
| FrontierMath Tiers 1–3 | Epoch AI Benchmarking Hub | 68.4% | – |
| FrontierMath Tier 4 | Epoch AI Benchmarking Hub | 21.9% | – |
| OTIS Mock AIME | Epoch AI Benchmarking Hub | 98.9% | – |
| ProofBench | Vals AI | 48.0%† | – |
| LiveBench Mathematics | LiveBench | 91.6% | – |
| Arena Maths | Arena (LMArena) | 1526 | – |
| Knowledge | |||
| SimpleQA Verified | Epoch AI Benchmarking Hub | 69.7% | – |
| LiveBench 2.0 Data Analysis | LiveBench | 54.0%† | – |
| LiveBench Research | LiveBench | 79.1%† | – |
| LiveBench 1.0 Data Analysis | LiveBench | 54.0% | – |
| AA-Omniscience v1.1 | Artificial Analysis | 23.3 | 53.5 |
| MMLU-Pro | Vals AI | 90.2% | – |
| LegalBench | Vals AI | 87.0% | – |
| TaxEval | Vals AI | 74.5% | – |
| PRBench Finance | Scale AI SEAL | 49.5%† | – |
| PRBench Legal | Scale AI SEAL | 48.9%† | – |
| BioMysteryBench | Vals AI | 62.2% | – |
| Excel Modeling Benchmark | Vals AI | 72.2% | – |
| Excel Modeling Benchmark v1.1 | Vals AI | 72.9% | 76.5% |
| MedCode | Vals AI | 48.1% | – |
| MedScribe | Vals AI | 84.5% | – |
| Public Benefits Bench v1.1 | Vals AI | 65.3% | – |
| GDP.pdf | Epoch AI Benchmarking Hub | 23.2% | – |
| AA-Omniscience v1.1: accuracy | Artificial Analysis | 50.5% | 58.3% |
| AA-Omniscience v1.1: non-hallucination | Artificial Analysis | 44.9% | 88.6% |
| Instruction following | |||
| LiveBench Language | LiveBench | 87.8% | – |
| LiveBench Instruction Following | LiveBench | 81.4% | – |
| Arena Instruction Following | Arena (LMArena) | 1490 | – |
| Human preference | |||
| Arena Text | Arena (LMArena) | 1496 | – |
| Arena Creative Writing | Arena (LMArena) | 1485 | – |
| Arena Multi-turn | Arena (LMArena) | 1504 | – |
| Arena Longer Queries | Arena (LMArena) | 1510 | – |
| Multimodal | |||
| Arena Vision | Arena (LMArena) | 1315 | – |
| LiveBench Design | LiveBench | 72.7%† | – |
| MMMU-Pro | Artificial Analysis | 85.6% | – |
| MMMU-Pro (Vals) | Vals AI | 89.1% | – |
| MortgageTax | Vals AI | 65.3% | – |
| SAGE | Vals AI | 88.2% | – |
| AudioMC | Scale AI SEAL | 60.4% | – |
| AudioMC (text output) | Scale AI SEAL | 60.4% | – |
| Blueprint-Bench 2 | Epoch AI Benchmarking Hub | 38.6%† | 58.6%† |
| AA-3D | Artificial Analysis | 966 | – |
| Long context | |||
| AA-LCR v1.1 | Artificial Analysis | 81.3% | 79.3% |
| MLCR | Artificial Analysis | 10.6% | – |
| Composite | |||
| Epoch Capabilities Index | Epoch AI Benchmarking Hub | 157.4† | – |
| LiveBench 2.0 | LiveBench | 59.2%† | – |
| LiveBench 1.0 | LiveBench | 75.8% | – |
| AA Intelligence Index v5.0 | Artificial Analysis | 30.6 | 59.4 |
| AA Intelligence Index v4.3.2 | Artificial Analysis | 40.9 | – |
| AA Progress Index v1 | Artificial Analysis | 320.3 | 351.6 |
| Vals Index | Vals AI | 56.0 | 70.4 |
| Vals Finance Index | Vals AI | 60.5% | 66.3% |
| Vals RSI Index | Vals AI | 19.9% | – |
| Vals Science Index | Vals AI | 35.8% | – |
| Vals Router Index | Vals AI | 56.0%† | – |
| AA Cyber Index | Artificial Analysis | 25.1 | – |
Data as of 2026-10-29. Best configuration of each model; every score links to its source on the model pages.
What changed
- Gemini 3.8 Flash: list price moves to $1.50 / $7.50 per 1M input / output tokens (announced).
- Gemini 3.8 Flash (high): BenchLeader Index 63.3 → 61.9.
- Gemini 3.8 Flash (high): 5 new results (AA Intelligence Index v5.0, AA Progress Index v1, HLE-Diamond (AA)…); 3 score revisions; cost per task moved from $3.49/M to $7.77/M (123% up).
- Gemini 3.8 Flash (high): cost per task moved from $1.24/M to $3.49/M (180% up).
- Gemini 3.8 Flash (high): 8 new results (Code Migration v1.1, Excel Modeling Benchmark v1.1, Finance Agent v2.1…).
- Gemini 3.8 Flash (high): 1 score revision.
- Gemini 3.8 Flash (high): 3 score revisions.
- Gemini 3.8 Flash (high): 1 new result (AudioMC (text output)).
- Gemini 3.8 Flash (high): 1 new result (SREBench); 2 score revisions.
- Gemini 3.8 Flash (high): 2 score revisions.
Related comparisons
- Gemini 3.8 Flash vs Claude Opus 5.5
- Gemini 3.8 Flash vs GPT-6 Astra
- Gemini 3.8 Flash vs GPT-6.1 Sol
- 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 3.7 Flash vs Gemini 4 Carbon
Rankings: Best LLM overall · Best LLM for coding · Best LLM for agents · Best cheap LLMs
Gemini 3.8 Flash vs Gemini 4 Carbon: questions
- Is Gemini 3.8 Flash better than Gemini 4 Carbon?
- Gemini 3.8 Flash (high) and Gemini 4 Carbon (high) are neck and neck on quality (61.8 vs 65.1, within the ±7.1 margin). The BenchLeader Index combines every independent quality benchmark; the two are within the margin of error as of 29 Oct 2026, so check the category scores for your use.
- Gemini 3.8 Flash vs Gemini 4 Carbon: which is cheaper, and will that change?
- Gemini 3.8 Flash is cheaper today: $1.50 against $8.00 per million tokens, blended at three input tokens per output token. Gemini 3.8 Flash's current price runs until 31 Dec 2026, then $3.00.
- Is Gemini 3.8 Flash or Gemini 4 Carbon better for coding?
- Gemini 4 Carbon scores higher in coding (72 vs 60 on the category index, where 50 is average). On FrontierCode 2.0 (AA), the top shared coding benchmark, it is 18.7% for Gemini 3.8 Flash against 40.6% for Gemini 4 Carbon.
- Which is faster, Gemini 3.8 Flash or Gemini 4 Carbon?
- Gemini 3.8 Flash streams faster (143 vs 100 output tokens per second, Artificial Analysis 7-day median).
- Which is better for long documents?
- Gemini 3.8 Flash scores higher on long-context benchmarks (66 vs 65). Both take about 1M tokens: Gemini 3.8 Flash 1,048,576, Gemini 4 Carbon 1,000,000.
More questions (3)
- Which has lower latency?
- Gemini 3.8 Flash answers first: 26.7 s against 103.1 s to the first answer token, as Artificial Analysis measures it. Reasoning effort drives most of this: a higher setting thinks longer before it answers.
- Do independent benchmarks agree on Gemini 3.8 Flash vs Gemini 4 Carbon?
- Across the 25 benchmarks both have, Gemini 3.8 Flash leads 1, Gemini 4 Carbon leads 24. Every overall index that rates both puts Gemini 4 Carbon ahead.
- When was Gemini 4 Carbon released?
- Gemini 4 Carbon was released on 28 Oct 2026.