Gemini 3.8 Flash
Google · released Sep 2, 2026
Gemini 3.8 Flash is Google's most intelligent Flash model with significant gains from 3.7 Flash across software engineering, agentic tasks, and multi-step reasoning.
Specification
- Context window
- 1.05M
- Max output
- 66K
- Knowledge cutoff
- Not stated
- Parameters
- Undisclosed
- Licence
- Proprietary
- Serving providers
- 2
- Moderated
- No
- Uptime
- 100.0%
Intelligence
40.9
92th percentile
Coding
76.3
Coding Index
Agentic
40.2
Agentic Index
Output speed
104 t/s
Median across providers
Latency
1.81s
Time to first token
Cost per task
$0.34
Estimated
Benchmarks
Where the score comes from
The Intelligence Index is a composite. These are the underlying evaluations this model was actually measured on.
Evaluation scores
Percentage correct · higher is better
- GPQA Diamond95.3%
- AA-LCR (long context)81.3%
- SciCode56.6%
- Humanity's Last Exam47.8%
- τ²-bench (Telecom)44.9%
An evaluation missing from this list was not run for this model — it is not a zero.
View as table
| Evaluation | Score |
|---|---|
| GPQA Diamond | 95.3% |
| AA-LCR (long context) | 81.3% |
| SciCode | 56.6% |
| Humanity's Last Exam | 47.8% |
| τ²-bench (Telecom) | 44.9% |
Against its peers
Intelligence Index · this model highlighted, nearest peers in grey
- GLM 5.3 FlashZ.ai41.8
- Claude Opus 4.8Anthropic41.8
- Nex-N2-ProNex AGI41.7
- Ling 3.1 FlashInclusionAI41.1
- 41.1
- Gemini 3.8 FlashGoogle40.9
- Claude Opus 4.7Anthropic40.7
- 39.9
Peers are the models sitting closest on the Intelligence Index — the set you would realistically choose between.
View as table
| Model | Intelligence |
|---|---|
| GLM 5.3 Flash | 41.8 |
| Claude Opus 4.8 | 41.8 |
| Nex-N2-Pro | 41.7 |
| Ling 3.1 Flash | 41.1 |
| Qwen3.6 Max Preview | 41.1 |
| Gemini 3.8 Flash | 40.9 |
| Claude Opus 4.7 | 40.7 |
| Qwen3.8 2.4T A95B | 39.9 |
Percentile among all indexed models
Pricing
What it costs to run
List prices per million tokens, plus what one representative task works out to.
List price
- Input / 1M tokens
- $0.75
- Output / 1M tokens
- $3.75
- Cached input / 1M
- $0.075
- Blended 3:1
- $1.5
One task, estimated
$0.34
- Input tokens
- 50,000
- Output tokens
- 80,000
- Profile
- Reasoning
Estimated from list pricing: 50K input tokens plus 80K output tokens for reasoning models (25K for non-reasoning).
Serving providers
Speed and latency figures are medians across these providers, so a widely-served model reports a blend rather than any single endpoint.
Arena
Head-to-head generation quality
Elo from pairwise judgements, broken out by the kind of thing the model was asked to build.
- Overall Elo
- 1,321
- Win rate
- 53.9%
- Strongest at
- UI components
- Tournaments
- 1,719
Elo by category
Dot position on a 1,225–1,325 scale · Elo has no meaningful zero
- 1,321
- 1,308
- 1,303
- 1,295
- 1,292
- 1,263
- 1,244
- 1,233
Agent categories (full-stack apps, mobile apps) are only scored for models tested in agent harnesses.
View as table
| Category | Elo | Win rate |
|---|---|---|
| UI components | 1321 | 53.9% |
| Game development | 1308 | 51.0% |
| Websites | 1303 | 50.0% |
| 3D scenes | 1295 | 49.9% |
| SVG | 1292 | 54.7% |
| Data visualisation | 1263 | 50.2% |
| Full-stack apps | 1244 | 48.1% |
| Mobile apps | 1233 | 54.4% |