Google launches Gemini 3.8 Flash for agents

// 2h agoMODEL RELEASE

Google launches Gemini 3.8 Flash for agents

Google’s Gemini 3.8 Flash is a cost-efficient model for long-horizon software engineering, autonomous agents, multimodal tasks, and complex reasoning. It supports a 1M-token context window, tool use, code execution, and computer use at Flash pricing.

// ANALYSIS

Gemini 3.8 Flash makes agentic coding more economically viable, though developers should validate Google’s benchmark claims against their own workloads.

–Targets autonomous, multi-step engineering tasks rather than simple code completion
–Maintains introductory pricing of $0.75 per million input tokens and $3.75 per million output tokens
–Adjustable reasoning effort lets developers trade off latency, quality, and token costs
–Supports function calling, search grounding, file search, code execution, and computer use preview
–Its 1M-token context window is well suited to large repositories and long-running agent sessions

// TAGS

gemini-3.8-flashllmai-codingcoding-agentagentmultimodallong-contexttool-use

DISCOVERED

2h ago

2026-09-02

PUBLISHED

3h ago

2026-09-02

RELEVANCE

10/ 10

AUTHOR

Prompt Engineering

// KEEP READING

More AI developer news from the feed

EXPLORE FULL FEED

INFRA55m ago

Anthropic signs reported $35B Lambda compute deal

Anthropic reportedly signed a $35 billion cloud-computing agreement with Nvidia-backed Lambda to secure future Nvidia GPU capacity at a Texas data center. The infrastructure will support rising demand for Claude and its increasingly autonomous workloads.

LAUNCH55m ago

Anthropic launches customer-controlled frontier safeguards

Anthropic’s Enterprise Frontier Safeguards combines cross-session misuse monitoring with customer-owned cloud storage, encryption keys, access policies, and audit controls. It supports Claude Code, Claude Enterprise, Claude Platform, Bedrock, Google’s Agent Platform, and Microsoft Foundry, with phased availability beginning later this fall.

OPEN SOURCE55m ago

DeepSeek V4 Vision Model Goes Open Source

DeepSeek has released downloadable weights for its 305-billion-parameter sparse multimodal model under the MIT license. The experimental checkpoint adds vision and a one-million-token context window, but self-hosting requires serious multi-GPU infrastructure and its benchmark claims remain unverified independently.