Gemini 3.8 Flash

What is Gemini 3.8 Flash?

Gemini 3.8 Flash builds on Gemini 3.7 Flash and is generally available on the Gemini API, in Google AI Studio and Antigravity, and for Google AI Pro and Ultra subscribers in the Gemini app. It reads text, images, video, audio and PDFs across a 1M-token window, writes up to 64K tokens, and its knowledge runs to March 2026 in some areas and January 2025 in others.

Its gains come from working harder: it takes extra reasoning steps and calls tools again before it answers, which lifts coding and agent scores but uses more tokens. Google keeps 3.7 Flash fully supported for jobs where token cost comes first.

What changed from Gemini 3.7 Flash?

Better coding and agent scores at the same list price, in exchange for more tokens per task.

Coding

73.7% on DeepSWE v1.1, up from 65.3%, and 89.4% on Terminal-Bench 2.1, up from 85.8%.

Knowledge work

1545 Elo on GDPval-AA v2 and 61.4% on Vals Finance Agent v2, up from 1482 and 59.0%.

Token use

It takes extra reasoning steps and calls tools more often, so it can use more tokens per task, especially at higher thinking levels.

Price

Unchanged: both list at $1.50 / $7.50 per million tokens, and 3.7 Flash stays fully supported.

Who is Gemini 3.8 Flash best for?

Gemini 3.8 Flash is best for long coding jobs, agents that run many steps on a budget, finance and legal analysis, and work on long documents, images and video.

• Long coding jobs

  Google tuned it for long-horizon software work, and it scored 73.7% on DeepSWE v1.1, within half a point of Claude Opus 5 in Google's table.

• Agents on a budget

  It is the default model behind Google's Antigravity agent, with thinking levels that trade tokens for quality, at $1.50 / $7.50 per million tokens.

• Finance and legal analysis

  It leads Google's table on Vals Finance Agent v2 at 61.4% and Harvey's Legal Agent Benchmark at 10.0%.

• Documents, images and video

  A 1M-token window takes text, images, video, audio and PDFs, and it scored 86.2% on CharXiv Reasoning and 87.8% on LVBench.

What's included with Gemini 3.8 Flash?

Three thinking levels, computer use in preview, Google Search and Maps grounding, built-in tools, the default Antigravity agent, and cheaper Batch and Flex tiers.

Three thinking levels

Low, medium (the default) and high trade speed and tokens against depth; asking for minimal returns an error.

Computer use

It can drive a computer interface through the Gemini API's computer use tool, which is in preview.

Google Search and Maps grounding

Paid use includes 5,000 grounded requests a month, shared across Gemini 3 models, then $14 per 1,000.

Built-in tools

Code execution, file search, URL context, function calling and structured outputs work out of the box.

Default Antigravity agent

Google's managed Antigravity agent and the Antigravity SDK now run on 3.8 Flash unless you pick another model.

Batch, Flex and Priority tiers

Half-price Batch and Flex for work that can wait, and a Priority tier for business-critical, low-latency traffic.

How much does Gemini 3.8 Flash cost?

Gemini 3.8 Flash is free within rate limits on the Gemini API's free tier; paid use costs $1.50 per million input tokens and $7.50 per million output tokens, or half that on Batch and Flex.

API prices are for the Gemini Developer API in US dollars. Priority costs $2.70 / $13.50 per million tokens. Paid API use needs a linked billing account and at least $5 of prepaid credit, which expires after 12 months; paid-tier prompts are not used to improve Google's products. The free Search grounding allowance is shared across all Gemini 3 models, and enterprise deals with provisioned throughput and volume discounts go through Google's sales team. The free Gemini app plan uses 3.6 Flash; Google AI Ultra, from $99.99 a month, also includes 3.8 Flash, at US prices. Full pricing on ai.google.dev. We checked these prices on 28 Sep 2026.

How do readers rate Gemini 3.8 Flash?

No one has rated Gemini 3.8 Flash yet. Be the first.

Gemini 3.8 Flash pros and cons

It is fast and matches Claude Opus 5 on Google's long-horizon coding test for far less, but it spends more tokens than 3.7 Flash and trails Opus badly on general agent tasks.

What are the best Gemini 3.8 Flash alternatives?

Five tools on Recaply do Gemini 3.8 Flash's job. DeepSeek V4.1 Flash, MiMo-V2.6-Flash and Step-3.5-Flash are open source, so you can run them on your own server.

Gemini 3.7 FlashFree

Choose it when token spend matters more than peak quality: it has the same list price, Google still fully supports it, and it uses fewer tokens per task.

Claude Opus 5.5Paid

Choose it when you need a stronger general agent or computer use: Google's own table puts 3.8 Flash far behind Claude Opus 5 there, and Anthropic reports Opus 5.5 ahead of Opus 5, at $4 / $20 per million tokens.

DeepSeek V4.1 FlashOpen Source

Choose it when price matters most: it costs $0.30 / $1.20 per million tokens at peak hours, half that off-peak, and can write up to 384K tokens in one answer.

MiMo-V2.6-FlashOpen Source

Choose it when you want a 309B omnimodal model for coding and agents under MIT that you can download.

Step-3.5-FlashOpen Source

Choose it when you want a fast open model with a 256K context for agents, self-hosted or on StepFun's API.

Questions about Gemini 3.8 Flash