Gemini 3.8 Flash: specs, pricing and benchmarks

Gemini 3.8 Flash specs

The table below lists the core specifications for gemini-3.8-flash on the Gemini API.

Google describes Gemini 3.8 Flash as its most intelligent Flash model, engineered for long-horizon software engineering, autonomous agents, and complex enterprise workflows. It takes text, image, video, audio, and PDF inputs and returns text.

The model is built on Gemini 3.7 Flash, which Google deprecated and rerouted to gemini-3.8-flash on October 8, 2026. Its knowledge cutoff is March 2026, though some domains are limited to January 2025.

Gemini 3.8 Flash pricing and free tier

The Gemini API free tier costs nothing for input, output, and context caching. On the paid tier, every Gemini 3.8 Flash token price doubles on January 1, 2027, as the table below shows, and Gemini API pricing covers the other main models.

• Prompts and responses on the free tier are used to improve Google products, while paid-tier content isn't, except on the invite-only Contributor tier.
• Batch, Flex, and Grounding with Google Search are only available on the paid tier, where Priority inference costs $1.35 in and $6.75 out per 1M tokens through December 31, 2026, then $2.70 and $13.50.

Gemini 3.8 Flash benchmarks

The numbers in the table below come from Google's own model card, comparing Gemini 3.8 Flash with Gemini 3.7 Flash, Claude Opus 5, and GPT-5.6 Sol.

• It beats Gemini 3.7 Flash on every benchmark in Google's table and scores highest of the four models on Vals Finance Agent v2, Harvey's Legal, Terminal-bench 2.1, CharXiv, HLE-Verified, and LABBench2.
• It trails Claude Opus 5 by a wide margin on Terminal-bench 4.0, OSWorld-2.0, and GDPVal-AA v2.

API features and limits

Thinking is built into the Gemini 3.8 Flash API with three supported levels: low, medium, and high. The default level is medium, and thinking tokens count toward your output token bill on the paid tier. Sending minimal as the thinking level isn't supported on gemini-3.8-flash and returns an error.

Both explicit and implicit context caching are supported. Caching is free on the free tier, while the paid tier bills a per-hour cache storage rate per 1M tokens on top of the cached input token price.

Supported tools and capabilities include code execution, computer use (in preview), file search, function calling, Grounding with Google Maps, Search grounding, structured outputs, and URL context. Output is text only, and audio generation, image generation, and the Live API aren't supported. Batch, Flex, and Priority inference all work on the paid tier.

Where Gemini 3.8 Flash runs

• Google AI Pro and Ultra subscribers can use it in the Gemini app and in AI Mode.
• Developers can call it through the Gemini API and in Google AI Studio on both the free and paid tiers.
• It's available for enterprise teams on the Gemini Enterprise Agent Platform.
• On Antigravity, it's the core agent model on every plan and the default model for the Antigravity agent in Managed Agents and the Antigravity SDK.
• All requests to Gemini 3.7 Flash have been rerouted to Gemini 3.8 Flash since October 8, 2026.
• Gemini 3.8 Flash Cyber is a separate cybersecurity version that Google makes available only to trusted defenders through its Fairwind Program.

Gemini 3.8 Flash vs Gemini 4 Argon

Gemini 4 Argon is Google's flagship model. It has no free tier and costs $2.00 for input and $10.00 for output per 1M tokens through November 30, 2026, compared with $0.75 in and $3.75 out for Gemini 3.8 Flash.

On DeepSWE v1.1, which both model cards report, Argon scores 77.9% against 73.7% for 3.8 Flash, and on Terminal-bench 4.0 it leads 57.4% to 19.1%.

Pick 3.8 Flash for high-volume work and access to the free tier, and Argon for the hardest reasoning and agent tasks.

Gemini 3.8 Flash through a Google AI plan

If you already pay for Google AI Pro or Ultra, Plan2API lets you use that Gemini subscription as an API and exposes the model as gemini/gemini-3.8-flash-high (google/ and antigravity/ prefixes work too) on an OpenAI-compatible Chat Completions endpoint (POST /v1/chat/completions) at https://api.plan2api.com/v1. You get a 1,048,576-token context window with support for text, image, audio, and video input.

Requests count against your Google AI plan's own usage limits, and Plan2API never bills per token. You can link your Google account at plan2api.com/connect to get a p2a_ key and send it in an Authorization: Bearer header. Here is a curl example:

curl https://api.plan2api.com/v1/chat/completions \
  -H "Authorization: Bearer $PLAN2API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "gemini/gemini-3.8-flash-high", "messages": [{"role": "user", "content": "Summarize this changelog in three bullets"}]}'