> ## Documentation Index
> Fetch the complete documentation index at: https://docs.tess.im/llms.txt
> Use this file to discover all available pages before exploring further.

# Gemini 3.8 Flash

Gemini 3.8 Flash is the most intelligent Flash model in Tess — built for **long-horizon coding, autonomous agents, and multi-step reasoning**, with **50% off through the end of 2026**.

| **Model ID**<br /><br />`gemini-3.8-flash`                                                   | **Context**<br /><br />1M input / 64K output | **Provider**<br /><br />Google DeepMind | **Released**<br /><br />2 Sep 2026            |
| :------------------------------------------------------------------------------------------- | :------------------------------------------- | :-------------------------------------- | :-------------------------------------------- |
| **Capabilities**<br /><br /><Icon icon="brain" /><Icon icon="image" /><Icon icon="wrench" /> | **Speed**<br /><br />High                    | **Cost**<br /><br />Low (intro)         | **Intelligence**<br /><br />Text + multimodal |

## What changed vs Gemini 3.7 Flash

|                               | Gemini 3.7 Flash                 | Gemini 3.8 Flash                                    |
| ----------------------------- | -------------------------------- | --------------------------------------------------- |
| Focus                         | Coding + agents + knowledge work | Long-horizon coding + autonomous agents             |
| Context                       | 1M / 64K                         | 1M / 64K                                            |
| DeepSWE (long-horizon coding) | 65.3%                            | **Outperforms 3.7 and most larger frontier models** |
| HLE-Verified                  | —                                | **54.9%**                                           |
| Tess credits / 100 tokens     | 0.036 in / 0.180 out             | **0.036 in / 0.180 out** (50% off)                  |
| Modalities                    | Text, image, audio, video        | Text, image, audio, video                           |

Native **reasoning** (effort **low / medium / high**, default **medium**), **tools** (function calling / MCP, search, computer use), and **vision**. On hard tasks it works harder — extra reasoning steps and iterative tool calls, which can use more tokens at higher effort.

## Pricing (Tess credits)

Values follow [Models and Costs](/en/models-and-cost) (credits per 100 tokens). This is the **50% off** rate through **31 December 2026**:

| Model            | Input / 100 tokens | Output / 100 tokens |
| ---------------- | ------------------ | ------------------- |
| Gemini 3.8 Flash | 0.036              | 0.180               |

<Tip>
  **50% off through the end of 2026.** Pricing changes from **1 January 2027**.
</Tip>

> <img src="https://mintcdn.com/tess-dfe1edf0/uqiv_DVeLCJNBSiH/images/Captura-de-Tela-2026-09-04-a%CC%80s-11.07.32.png?fit=max&auto=format&n=uqiv_DVeLCJNBSiH&q=85&s=011b9659e751f90192371ada2ad28f31" alt="Captura De Tela 2026 09 04 Às 11 07 32" width="1480" height="862" data-path="images/Captura-de-Tela-2026-09-04-às-11.07.32.png" />

## Ideal use cases in Tess

1. Long-horizon coding agents that have to finish a change end to end
2. Multi-step agent workflows — planning, tool loops, error recovery
3. Quantitative and professional analysis (finance, legal-style research)
4. Hard multi-step reasoning across STEM and knowledge work (HLE-Verified 54.9%)
5. Web development against visual references
6. Multimodal turns (text + image + audio + video)

<Tip>
  **Best practices**

  * Prefer **3.8 Flash** over 3.7 Flash when coding quality and agent reliability matter more than token count.
  * Keep **3.7 Flash** for efficiency-first workloads — Google still supports it, and 3.8 can spend more tokens at higher effort.
  * Keep **3.5 Flash-Lite** for high-volume extraction, classification, and summarization.
</Tip>

See also: [Gemini 3.7 Flash](/en/gemini-3-7-flash) · [Gemini Flash](/en/gemini-flash) · [Models and Costs](/en/models-and-cost) · [Gemini 3.8 Flash (Google)](https://blog.google/innovation-and-ai/models-and-research/gemini-models/3-8-flash-and-3-8-flash-cyber/).