New LLMs for Agents and Media Pipelines: Pricing, Safety

Today’s developer update focuses on new model releases that add capabilities for agentic workflows and generative media pipelines, along with pricing and platform availability details. Anthropic shipped Claude Sonnet 5 for tool-using, planning-style agents, including updated pricing and default cyber safeguards. Google DeepMind released Nano Banana 2 Lite for fast image generation and Gemini Omni Flash for video generation and conversational editing, both available through developer platforms.

1. Claude Sonnet 5 is now available across Anthropic plans and via the Claude API

• Claude Sonnet 5 can make plans, use tools like browsers and terminals, and run autonomously.
• Sonnet 5 is available from today as the default model for Free and Pro plans, and is available to Max, Team, and Enterprise users.
• Sonnet 5 pricing is $2 per million input tokens and $10 per million output tokens through August 31, 2026, then $3 per million input tokens and $15 per million output tokens, and developers can use it via the Claude API.
  Impact: Developers can deploy a lower-priced agentic model for multi-step tool use across Claude plans and API workflows.

2. Sonnet 5 includes updated safety posture and default cybersecurity safeguards

• Safety evaluations found an overall lower rate of undesirable behaviors than Sonnet 4.6 and lower ability to perform cybersecurity tasks than current Opus models.
• For cybersecurity tasks, Sonnet 5 can perform some routine, non-harmful cyber tasks, but shows substantially poorer performance on potentially dangerous cyber skills.
• Sonnet 5 is launched with cyber safeguards enabled by default that detect and block dangerous cyber usage in real time, with the safeguards described as less strict than those launched with Fable 5.
  Impact: Agentic deployments using Sonnet 5 can rely on real-time cyber task blocking being enabled by default.

3. Nano Banana 2 Lite and Gemini Omni Flash expand generative media model access for developers

• Nano Banana 2 Lite (gemini-3.1-flash-lite-image) is described as the fastest, most cost-efficient Nano Banana image model and is available in Google AI Studio, Gemini API, and Gemini Enterprise Agent Platform.
• Gemini Omni Flash is now available in Google AI Studio and the Gemini API for video generation and conversational editing, with support for multimodal inputs including text, image, and video.
• Nano Banana 2 Lite is positioned for fast text-to-image outputs in 4 seconds, and Gemini Omni Flash is priced at $0.10 per second of video output.
  Impact: Developers can build higher-throughput image-to-video pipelines by accessing fast image generation and conversational video editing via the Gemini API.

4. Gemini Omni Flash supports up to 10-second generations and multi-turn editing via Interactions API

• Gemini Omni Flash offers 10-second video generations currently, with longer durations described as coming soon.
• Uploading audio references and scene extension is described as not supported in the Gemini API for this model, and video references up to 3 seconds are accepted by the API schema but not correctly processed at this time.
• Developers can maintain session history and context using the Interactions API, with support for stacking up to three sequential edits.
  Impact: Video editing workflows can use multi-turn sessions with short (current) clips and structured edit chaining through the Interactions API.

Sources

• AnthropicNews: https://www.anthropic.com/news/claude-sonnet-5
• GoogleDeepmind: https://deepmind.google/blog/start-building-with-nano-banana-2-lite-and-gemini-omni-flash/