← Back to feed

Google’s Gemini 3.8 Flash launched on OpenRouter around September 2, 2026 at $0.75 per million input tokens and $3.75 per million output, with a context window of 1,048,576 tokens. The Gemini 3.8 Live voice model followed two weeks later on September 15, reaching the top position on Artificial Analysis’s Speech-to-Speech Quality Index.

Benchmark Placement

Gemini 3.8 Flash has been evaluated on LiveBench, where it scores 75.8 overall. Third-party harnesses have also run it on SWE-bench Verified and Terminal-Bench evaluations. It enters the 3.8 generation with real benchmark data available from launch, unlike some earlier mid-tier releases.

Pricing in Context

At $0.75/$3.75 per million, Gemini 3.8 Flash sits above the Flash Lite tier — Gemini 3.1 Flash Lite launched at $0.25/M input — and carries the same 1,048,576-token context window that Google has shipped across its Flash and Flash Lite lines. The output-to-input price ratio of 5:1 matches recent Google Flash releases.

The 3.8 Generation

Gemini 3.8 Flash arrived first in the 3.8 series. Two weeks later, Gemini 3.8 Live landed via the Google Live API on September 15, scoring 82.6 on Artificial Analysis’s Speech-to-Speech Quality Index to take first place among speech-to-speech models. Third-party tau-Voice benchmarks put Gemini 3.8 Live at 68.6% overall and 35.1% on Sierra’s banking-sector voice tasks.

Together, the two models cover standard text inference and real-time voice in the same generational tier.

Key Numbers

• Input cost: $0.75 per million tokens
• Output cost: $3.75 per million tokens
• Context window: 1,048,576 tokens
• LiveBench overall: 75.8
• Launch date: September 2, 2026
• Sibling: Gemini 3.8 Live (launched September 15, AA Speech-to-Speech #1 at 82.6)