Gemini 3.8 Flash
Google · Released Sep 2, 2026
$0.75
Input price / MTok
#10 of 33$3.75
Output price / MTok
#10 of 331.0M
Context window
#16 of 3473.7
DeepSWE v1.1
#1 of 189.4
Terminal-bench 2.1
#1 of 159
OSWorld-2.0
#1 of 161.4
Vals Finance Agent v2
#1 of 154.9
HLE-Verified
#1 of 1Summary
Gemini 3.8 Flash is a proprietary language model from Google, released on Sep 2, 2026.
At $0.75 input and $3.75 output per million tokens, it ranks #10 of 33 for input price, placing it mid-range on input cost.
Its context window of 1.0M tokens ranks #16 of 34 for context depth in this dataset.
It targets coding, agents, budget workloads.
Built on the Gemini 3.7 Flash base rather than a new foundation model, per Google's own model card. Tuned specifically for long-horizon coding and autonomous agents. Shipped one day before Claude Fable 5.1 and GPT-6 Astra, in a 48-hour window where three frontier labs released within days of each other; positioned as the cost-efficient option of the three rather than a raw-capability leader. Knowledge cutoff March 2026 (some domains limited to January 2025). Available same-day in GitHub Copilot.
Benchmarks
| Benchmark | Score | Max | % | Rank | Source |
|---|---|---|---|---|---|
| DeepSWE v1.1 | 73.7 | 100 | 73.7% | only entry | Google DeepMind model card (self-reporte... |
| Terminal-bench 2.1 | 89.4 | 100 | 89.4% | only entry | Google DeepMind model card (self-reporte... |
| OSWorld-2.0 | 59 | 100 | 59.0% | only entry | Google DeepMind model card (self-reporte... |
| Vals Finance Agent v2 | 61.4 | 100 | 61.4% | only entry | Google DeepMind model card (self-reporte... |
| HLE-Verified | 54.9 | 100 | 54.9% | only entry | Google DeepMind model card (self-reporte... |