Data as of October 17, 2026 · How the score is built
Released Sep 2, 20261M contextGoogle DeepMind Gemini 3.8 Flash model card
Gemini 3.8 Flash
Decision readingGemini 3.8 Flash scores 69.3 out of 100 and ranks #22 of 224. This profile shows 31 source-displayable benchmark rows; its strongest eligible category is Multimodal & Grounded at #9. API pricing is $0.75 input and $3.75 output per million tokens, with cached input at $0.075.
Released Sep 2, 2026 — see all recent releases
Decision snapshot
Each value carries a field reference instead of floating alone. Markers compare this model with the current ranked and priced catalog; they are not absolute quality thresholds.
Capability
69.3/100
field median 49.6#22 of 224 ranked models
Public
#22of 224
Verified #17 of 88
Price
$0.75input / $3.75 output
input median $0.75cached $0.075 · blended $2.25
Speed
147tok/s
field median 92 tok/sFirst token 23.8 s
Context
1Mtokens
field median 256,000Maximum output length is tracked separately
Strongest published evidence
Multimodal & Grounded ranks #9. Particularly strong for screenshots, documents, charts, and grounded multimodal workflows.
Validate before choosing
31 published rows leave some tracked benchmark slots empty. Reasoning is its lowest eligible category at #17.
Source-linked · 31 displayable benchmark rows
Category score record
Scores and ranks appear only where published evidence can be displayed. The table keeps the score, weight, cohort, and evidence state together.
| Category | Score | Rank | Percentile | Weight | Benchmarks | Evidence |
|---|---|---|---|---|---|---|
| AgenticRank #16 of 128Percentile 88thWeight 22%9 benchmarksVerified | 65.9 | 9 benchmarks | Verified | |||
| CodingRank #17 of 150Percentile 89thWeight 20%8 benchmarksVerified | 61.0 | 8 benchmarks | Verified | |||
| ReasoningRank #17 of 28Percentile 41stWeight 17%3 benchmarksVerified | 74.3 | 3 benchmarks | Verified | |||
| MultimodalRank #9 of 54Percentile 85thWeight 12%2 benchmarksVerified | 87.3 | 2 benchmarks | Verified | |||
| KnowledgeRank #13 of 184Percentile 93rdWeight 12%6 benchmarksVerified | 73.5 | 6 benchmarks | Verified | |||
| MultilingualWeight 7%0 benchmarksNot measured | Not measured | 0 benchmarks | Not measured | |||
| Inst. FollowingWeight 5%1 benchmarkVerified | Score pending | 1 benchmark | Verified | |||
| MathRank Not rankedWeight 5%2 benchmarksVerified | 69.7 | 2 benchmarks | Verified |
31 of 681 tracked benchmark slots have displayable evidence · bars run 0–100
Coverage detailsHow much of this is verified
Coverage is split by category so a strong number never hides a thin evidence base. Verified means the row is tied to a published source; provisional rows remain visible but separate.
- Agentic9/9 verified
- Coding8/8 verified
- Reasoning3/3 verified
- Multimodal2/2 verified
- Knowledge6/6 verified
- MultilingualNot measured
- Inst. Following1/1 verified
- Math2/2 verified
Capability shape
Each axis shows percentile within that category’s eligible cohort. The comparison outline is the median of the six nearest public-score peers; a collapsed vertex means the category is not rank-eligible.
Gemini 3.8 Flash category percentile values
- Agentic88th percentile
- Coding89th percentile
- Reasoning41st percentile
- Multimodal85th percentile
- Knowledge93rd percentile
- MultilingualNot eligible
- Instruction followingNot eligible
- MathNot eligible
The dashed outline is median of 6 nearest peers.
Eligible category ranks
- Agentic#16/128
- Coding#17/150
- Reasoning#17/28
- Multimodal#9/54
- Knowledge#13/184
- MultilingualNot ranked
- Inst. FollowingNot ranked
- MathNot ranked
Benchmark ledger
Coding opens by default. The marker compares each value with the best source-verified result in the catalog; provisional leaders do not set the reference. Expand the remaining categories for every published row.
Coding8 rows
| Benchmark | Score | Versus best verified row | Gap | Weight | Evidence |
|---|---|---|---|---|---|
| DeepSWE | Score73.8% | Versus best verified row Best verified: Muse Wonder Full · 79.0% | Gap5.2 behind | Weight15% ref. weight | Benchmark exact |
| CursorBench 3.2 | Score69.2% | Versus best verified row Best verified: Claude Fable 5.1 · 73.4% | Gap4.2 behind | Weight10% ref. weight | Benchmark exact |
| LiveCodeBench (Vals)LiveCodeBench, Vals AI run | Score89.5% | Versus best verified row Best verified: Claude Fable 5.1 · 90.5% | Gap1 behind | Weight8% ref. weight | |
| FrontierSWE v2 | Score19.6% | Versus best verified row Best verified: GPT-6 Astra · 65.5% | Gap45.9 behind | Weight8% ref. weight | Benchmark exact |
| Terminal-Bench 2.1 | Score89.4% | Versus best verified row Best verified: SWE-2 · 92.8% | Gap3.4 behind | WeightScored in Agentic | Provider exact |
| SWE-bench (Vals)SWE-bench, Vals AI run | Score80.0% | Versus best verified row Best verified: Claude Opus 5 · 97.0% | Gap17 behind | WeightDisplay only | Verified |
| CursorBench 4.0 | Score39.6% | Versus best verified row Best verified: Claude Opus 5.5 · 57.8% | Gap18.2 behind | WeightDisplay only | Benchmark exact |
| Bug Hunt Bench | Score18.0 fixes | Versus best verified row Best verified: Claude Sonnet 5.5 · 51.3 fixes | Gap33.3 behind | WeightDisplay only | Benchmark exact |
Agentic9 rows
| Benchmark | Score | Versus best verified row | Gap | Weight | Evidence |
|---|---|---|---|---|---|
| Terminal-Bench 2.1 | Score89.4% | Versus best verified row Best verified: SWE-2 · 92.8% | Gap3.4 behind | Weight8% ref. weight | Provider exact |
| Terminal-Bench 4.0 | Score19.10% | Versus best verified row Best verified: Claude Sonnet 5.5 · 70.60% | Gap51.5 behind | Weight8% ref. weight | Provider exact |
| Terminal-Bench 2.1 (Vals)Terminal-Bench 2.1, Vals AI run | Score81.3% | Versus best verified row Best verified: GPT-6 Astra · 87.3% | Gap6 behind | Weight4% ref. weight | |
| Finance Agent v2 | Score61.4% | Versus best verified row Best verified: Gemini 4 Argon · 65.4% | Gap4 behind | WeightDisplay only | Provider exact |
| OSWorld 2.0 | Score59.0% | Versus best verified row Best verified: GPT-6 Astra · 72.6% | Gap13.6 behind | WeightDisplay only | Provider exact |
| ApprenticeBenchApprenticeBench: end-to-end computer use, continual learning, and long-horizon agency on a real accounts-payable job | Score24% | Versus best verified row Best verified: Claude Fable 5.1 · 72% | Gap48 behind | WeightDisplay only | Benchmark exact |
| Hermes IndexHermes Index: agentic capability and cost per task inside Hermes Agent | Score38.1% | Versus best verified row Best verified: Claude Opus 5.5 · 63.3% | Gap25.2 behind | WeightDisplay only | Benchmark exact |
| Hermes BenchHermes Bench: 150 agent tasks graded on the files and state the agent leaves behind | Score62.2% | Versus best verified row Best verified: Claude Opus 5.5 · 76.7% | Gap14.5 behind | WeightDisplay only | Benchmark exact |
| MazeBenchMazeBench: visual spatial reasoning in a 3D open world | Score4% | Versus best verified row Best verified: GPT-6 Astra · 23% | Gap19 behind | WeightDisplay only | Benchmark exact |
Reasoning3 rows
| Benchmark | Score | Versus best verified row | Gap | Weight | Evidence |
|---|---|---|---|---|---|
| ARC-AGI-2Abstraction and Reasoning Corpus for AGI v2 | Score89.2% | Versus best verified row Best verified: GPT-6 Astra · 95% | Gap5.8 behind | WeightWeighted 25% | Benchmark exact |
| ARC-AGI-3Abstraction and Reasoning Corpus for AGI v3 | Score10.4% | Versus best verified row Best verified: GPT-6 Astra · 62.7% | Gap52.3 behind | WeightWeighted 15% | Benchmark exact |
| ARC-AGI-1ARC-AGI-1 Semi-Private Evaluation | Score98.50% | Versus best verified row Best verified: GPT-6 Astra · 98.50% | GapBest verified | WeightDisplay only | Benchmark exact |
Multimodal2 rows
| Benchmark | Score | Versus best verified row | Gap | Weight | Evidence |
|---|---|---|---|---|---|
| CharXiv w/o toolsCharXiv Reasoning without tools | Score86.2% | Versus best verified row Best verified: Claude Mythos 5 · 88.9% | Gap2.7 behind | WeightDisplay only | Provider exact |
| LVBench | Score87.1% | Versus best verified row Best verified: Gemini 4 Argon · 91.7% | Gap4.6 behind | WeightDisplay only | Provider exact |
Knowledge6 rows
| Benchmark | Score | Versus best verified row | Gap | Weight | Evidence |
|---|---|---|---|---|---|
| MMLU-Pro (Vals)MMLU-Pro, Vals AI run | Score90.2% | Versus best verified row Best verified: Claude Fable 5.1 · 92.4% | Gap2.2 behind | Weight6% ref. weight | Verified |
| GPQA Diamond (Vals)GPQA Diamond, Vals AI run | Score94.4% | Versus best verified row Best verified: Gemini 3.1 Pro · 95.5% | Gap1.1 behind | Weight2% ref. weight | |
| HLE-Verified | Score54.9% | Versus best verified row Best verified: Gemini 3.8 Flash · 54.9% | GapBest verified | WeightDisplay only | Provider exact |
| LABBench2LABBench2: An Improved Benchmark for AI Systems Performing Biology Research | Score86.2% | Versus best verified row Best verified: Gemini 4 Argon · 88.8% | Gap2.6 behind | WeightDisplay only | Provider exact |
| BioMysteryBench (human-solvable)BioMysteryBench Human Solvable | Score88.8% | Versus best verified row Best verified: Claude Opus 5 · 90.1% | Gap1.3 behind | WeightDisplay only | Provider exact |
| BioMysteryBench (human-difficult)BioMysteryBench Human Difficult | Score56.5% | Versus best verified row Best verified: Gemini 3.8 Flash · 56.5% | GapBest verified | WeightDisplay only | Provider exact |
Inst. Following1 row
| Benchmark | Score | Versus best verified row | Gap | Weight | Evidence |
|---|---|---|---|---|---|
| Gray Swan IPI (15 attempts)Gray Swan Indirect Prompt Injection, attack success at 15 attempts | Score5.5% | Versus best verified row Best verified: Gemini 4 Argon · 0.7% | Gap4.8 behind | WeightDisplay only | Provider exact |
Math2 rows
| Benchmark | Score | Versus best verified row | Gap | Weight | Evidence |
|---|---|---|---|---|---|
| FrontierMath v2 (Tiers 1-3)FrontierMath v2 Tiers 1-3 | Score68.421% | Versus best verified row Best verified: GPT-6 Astra · 93.684% | Gap25.3 behind | WeightWeighted 30% | Benchmark exact |
| FrontierMath v2 (Tier 4)FrontierMath v2 Tier 4 | Score21.951% | Versus best verified row Best verified: GPT-6.1 Sol · 100.000% | Gap78 behind | WeightWeighted 10% | Benchmark exact |
What it costs to get this score
Published API price against the public score. The x-axis uses a log scale; the dashed path marks models that are not beaten by a cheaper, higher-scoring option. Price uses average of published input and output rates.
Gemini 3.8 Flash · 69.3 score · $2.25 blended per million tokens
The chart opens on the current model. Scroll horizontally to inspect the full price axis.
Horizontal: blended price per million tokens, log scale · Vertical: public score
Lineage
The sequence follows explicit supersedes links. Each score is estimated for that model; a relative can inform a sparse estimate but never sets a floor, so a newer release can score below an earlier one. Scores and prices remain blank when the corresponding public row or first-party rate is unavailable.
Base entry
Gemini 3.8 Flash CyberSpec sheet
Each documented value carries its source. Missing fields stay visible as not sourced or not published, rather than disappearing from the page.
- API model ID
- gemini-3.8-flashGoogle Gemini 3.8 Flash API documentation
- Context window
- 1MGoogle Gemini 3.8 Flash API documentation
- Maximum output
- Not sourced yet
- Knowledge cutoff
- Not sourced yet
- Input modalities
- text, image, video, audio, pdfGoogle Gemini 3.8 Flash API documentation
- Output modalities
- textGoogle Gemini 3.8 Flash API documentation
- Parameters
- Not disclosed by the provider
- Availability
- Gemini API · Google AI Studio · Android Studio · Stitch · Gemini app · Gemini Enterprise · Google AI Mode · Google Sheets · Google AntigravityGoogle Gemini 3.8 Flash and Gemini 3.8 Flash Cyber launch
- Cloud regions
- Not tracked yet
- Lifecycle
- activeGoogle Gemini 3.8 Flash API documentation
- API capabilities
- Tool calling, structured outputs, and batch support are not tracked yet
- Prompt caching
- Published at $0.075 per million cached input tokensGoogle Gemini API pricing
- Self-host
- Weights are not published
- Rate limits
- Not tracked yet
How to read this profile
The visual layer above carries the decisions. These notes preserve the model, ranking, coverage, and family context behind the numbers.
Gemini 3.8 Flash ranks #22 of 224 on the public leaderboard with a score of 69.25/100. Its source-verified position is #17 of 88.
Gemini 3.8 Flash is a proprietary model with a 1M context window. It uses an explicit reasoning mode, which can improve complex problem solving while adding latency and token use.
Generally available as the stable `gemini-3.8-flash` model through the Gemini API and Google AI Studio. Google also distributes it through Android Studio, Stitch, the Gemini app, Gemini Enterprise, Google AI Mode, Google Sheets, and Google Antigravity.
Gemini 3.8 Flash sits in the Gemini 3.8 Flash family with Gemini 3.8 Flash Cyber. Its explicit predecessor is Gemini 3.7 Flash. 31 of 681 tracked benchmark slots currently have displayable evidence. Missing categories stay blank.
Its strongest eligible category is Multimodal & Grounded at #9, while its lowest eligible position is Reasoning at #17. particularly strong for screenshots, documents, charts, and grounded multimodal workflows.
Last updated October 17, 2026. Runtime fields remain blank until a sourced snapshot exists.
Questions
How does Gemini 3.8 Flash perform overall in AI benchmarks?
Gemini 3.8 Flash ranks #22 out of 224 models on the public BenchAlign leaderboard, with a score of 69.25/100. Its evidence status is Supported, and this profile shows 31 source-displayable benchmark rows. The label describes evidence depth, not a provider quality claim; inspect category rows before choosing a workload.
Is Gemini 3.8 Flash good for knowledge and understanding?
Gemini 3.8 Flash ranks #13 out of 184 eligible models for knowledge and understanding, with a public category score of 73.5/100. Higher-ranked alternatives are available for workloads where this category decides the choice. Check the underlying rows before treating the aggregate as a workload guarantee.
Is Gemini 3.8 Flash good for coding and programming?
Gemini 3.8 Flash ranks #17 out of 150 eligible models for coding and programming, with a public category score of 61/100. Higher-ranked alternatives are available for workloads where this category decides the choice. Check the underlying rows before treating the aggregate as a workload guarantee.
Is Gemini 3.8 Flash good for mathematics?
Gemini 3.8 Flash has source-displayable benchmark coverage for mathematics, but the public category table does not assign it a rank there. The individual rows remain available for inspection. A missing category position means the evidence threshold was not met; it does not convert the model's unmeasured work into a zero.
Is Gemini 3.8 Flash good for reasoning and logic?
Gemini 3.8 Flash ranks #17 out of 28 eligible models for reasoning and logic, with a public category score of 74.3/100. Higher-ranked alternatives are available for workloads where this category decides the choice. Check the underlying rows before treating the aggregate as a workload guarantee.
Is Gemini 3.8 Flash good for agentic tool use and computer tasks?
Gemini 3.8 Flash ranks #16 out of 128 eligible models for agentic tool use and computer tasks, with a public category score of 65.9/100. Higher-ranked alternatives are available for workloads where this category decides the choice. Check the underlying rows before treating the aggregate as a workload guarantee.
Is Gemini 3.8 Flash good for multimodal and grounded tasks?
Gemini 3.8 Flash ranks #9 out of 54 eligible models for multimodal and grounded tasks, with a public category score of 87.3/100. That places it in the current top ten for this category. Check the underlying rows before treating the aggregate as a workload guarantee.
Is Gemini 3.8 Flash good for instruction following?
Gemini 3.8 Flash has source-displayable benchmark coverage for instruction following, but the public category table does not assign it a rank there. The individual rows remain available for inspection. A missing category position means the evidence threshold was not met; it does not convert the model's unmeasured work into a zero.
Which sibling models are related to Gemini 3.8 Flash?
Gemini 3.8 Flash belongs to the Gemini 3.8 Flash family. Related tracked variants include Gemini 3.8 Flash Cyber. A sibling link indicates shared lineage or a documented configuration relationship; it does not mean the variants have identical pricing, context limits, benchmark evidence, or deployment behavior. Compare before switching.
Does Gemini 3.8 Flash have full benchmark coverage on BenchLM?
No. Gemini 3.8 Flash currently has 55 source-displayable rows across 681 tracked benchmark slots. The profile exposes published, non-generated evidence and leaves missing categories blank until an exact evaluation is available. Coverage describes how much was measured; it is not a penalty added to an individual benchmark result.
What is the context window size of Gemini 3.8 Flash?
Gemini 3.8 Flash has a documented context window of 1M. That figure is the maximum combined prompt and retained-conversation space reported for this exact model; it is not the maximum output length. The profile keeps output limits separate because providers often publish those limits independently.