Gemini 3.8 Flash - Kyma API

Overview

Gemini 3.8 Flash is Google’s September 2026 Flash-tier release. It accepts text, image, audio, video and file input and returns text, with reasoning, tool calling and structured outputs on one request. Google’s model card puts the knowledge cutoff at March 2026. Against 3.7 Flash the shape is unchanged — 1,048,576 tokens of context and 65,536 tokens of output — and the difference is quality: Google positions it as its most intelligent Flash model, with the biggest gains on software engineering, long-horizon agentic work and multimodal understanding. On Kyma it serves through an OpenAI-compatible endpoint with automatic failover to a second independent route. Prompt caching is supported and repeated prefixes bill at this model’s cached input rate. Every response carries the exact request cost in usage.cost and names the model that answered in the X-Kyma-Model header.

Specs

Field	Value
Model ID	`gemini-3.8-flash`
Best for	High-volume agentic and multimodal work at 1M context
Context window	1.048576M
Max output tokens	66K
Input modalities	Text, Image, Video, File, Audio
Output modalities	Text
Tool calling	Yes
Structured outputs	Yes
Prompt caching	Yes
Speed	Fast
Cost band	Premium
Release stage	Stable

Pricing

	Per 1M tokens
Input	$2.026
Output	$10.126

Use this when

Agentic coding at Flash cost — Multi-step tool use over a large repository, with reasoning and structured outputs, at the Flash tier’s price and latency.
Long-form structured extraction — Reads a 1M-token corpus and emits tens of thousands of tokens of schema-constrained JSON in a single response.
Whole-document translation and rewriting — The 64K output ceiling fits an entire long document’s rewritten text without splitting the job into chunks.
Multimedia review — Takes video, audio, images and files alongside a text prompt and returns a long written analysis in one pass.

Not ideal for

Do not reach for this when a short reply is all you need and cost is the deciding factor — the cheaper Flash tiers answer the same short prompts, and the large output ceiling buys nothing you will use.

Pick something else when

You need the cheapest 1M-context tier and replies stay short: use gemini-3.5-flash-lite.
You want the previous Flash generation at the same list rate: use gemini-3.7-flash.
You need a frontier-tier model rather than a Flash tier: use claude-fable-5.1 or gpt-5.6-sol.

Example

from openai import OpenAI

client = OpenAI(base_url="https://kymaapi.com/v1", api_key="kyma-...")

response = client.chat.completions.create(
    model="gemini-3.8-flash",
    messages=[{"role": "user", "content": "..."}],
    tools=[...],  # function calling supported
)

Agent query example

Ask the API which models fit, instead of hardcoding an id:

curl "https://kymaapi.com/v1/models?recommended_for=chat&tools=true&quality_tier=frontier-open"

FAQ

How is this different from Gemini 3.7 Flash? Same context (1M tokens), same output ceiling (64K), same input set and same list price. The difference is the model itself: Google reports its largest Flash-to-Flash gains on software engineering and agentic tasks. If 3.7 Flash already does the job, nothing forces a move; new work should start on 3.8. Is the current price permanent? No. Google publishes a launch rate through December 31, 2026 and a higher list rate after that; the Pricing table on this page shows the promotional rate, the date it is promised through, and the list rate it returns to. Does prompt caching work on this model? Yes. Repeated prompt prefixes bill at this model’s cached input rate, and the saving is reported per request in usage.cache_discount.

​Overview

​Specs

​Pricing

​Use this when

​Not ideal for

​Pick something else when

​Example

​Agent query example

​FAQ