Latest AI Model Releases: GPT, Claude & Gemini

AI model names change quickly, but the useful question stays the same: what is actually new, where can you use it, and is it worth switching? This tracker covers significant general-purpose model releases from the largest AI labs—not every minor app update or experimental checkpoint.

Current as of July 13, 2026. We verify launch status against first-party announcements. Preview models and restricted-access models are labeled so they are not confused with broadly available releases.

The latest releases at a glance

What changed in the newest models

OpenAI: GPT-5.6 Sol, Terra, and Luna

OpenAI released the GPT-5.6 family on July 9. Instead of one model with a few size suffixes, it uses three capability tiers:

• Sol is the flagship for the hardest coding, research, and professional tasks.
• Terra balances capability and cost for everyday production work.
• Luna prioritizes speed and price for high-volume workloads.

All three are available through the OpenAI API as well as OpenAI's products. The API also adds programmatic tool calling, persisted reasoning, explicit prompt-cache controls, and beta multi-agent orchestration. OpenAI lists standard API prices per one million tokens at $5 input / $30 output for Sol, $2.50 / $15 for Terra, and $1 / $6 for Luna.

Try it if: you want one provider with clear model tiers, strong tool use, and access across chat, coding, and API workflows.

Read OpenAI's GPT-5.6 announcement or compare costs in our LLM API pricing guide.

Anthropic: Claude Sonnet 5 and Fable 5

Anthropic's current lineup separates a scalable workhorse from its highest-capability general model.

Claude Sonnet 5, announced June 30, targets coding, agents, and professional work at scale. It is the sensible starting point for most Claude users because it is designed for sustained production use rather than only the hardest edge cases.

Claude Fable 5 is Anthropic's more capable option for difficult software engineering, research, vision, and knowledge work. Anthropic initially launched it June 9, temporarily suspended access, and made it available again July 1. Its API price is $10 per million input tokens and $50 per million output tokens. Claude Mythos 5 uses the same underlying model with different safeguards, but remains restricted to approved programs; it should not be treated as a generally available alternative.

Try Sonnet 5 if: you need a default Claude model for coding or agent workflows. Move to Fable 5 if: task quality matters more than latency and cost, and the work is complex enough to justify the premium.

See Anthropic's release newsroom and Fable 5 availability details.

Google: Gemini 3.5 Flash

Google introduced Gemini 3.5 with Gemini 3.5 Flash on May 19. The release focuses on long-horizon agentic work, coding, multimodal understanding, and fast output. It is available in the Gemini app and AI Mode, through the Gemini API in Google AI Studio, and in Google's enterprise products.

Google said at launch that Gemini 3.5 Pro was still in development. Until Google announces broad availability, treat Flash as the current released member of the 3.5 family rather than assuming Pro is production-ready.

Try it if: speed, multimodal input, or high-throughput agent tasks matter more than selecting the largest possible model.

Read Google's Gemini 3.5 announcement.

Which new model should you use?

There is no durable “best model” for every workload. Start with the cheapest model that can pass your own evaluation set, then move up only when failures justify the extra cost.

For a production decision, test 20–50 representative prompts and score correctness, latency, tool-call reliability, and total cost. Public benchmarks can narrow a shortlist, but they cannot reproduce your data, prompts, or failure tolerance.

How we decide what belongs on this page

We add a release when it materially changes capability, cost, availability, or the recommended default for real users. We distinguish four labels:

• Generally available: broadly usable in a product or public API.
• Preview: usable, but behavior or terms may change before general availability.
• Restricted: limited to selected partners, researchers, or safety programs.
• Announced: described publicly but not yet available to typical users.

That distinction matters. A model can lead a benchmark and still be the wrong recommendation if developers cannot reliably access it.

Frequently asked questions

What is the latest GPT model?

The latest major OpenAI release is the GPT-5.6 family, released July 9, 2026. Sol is the flagship, Terra is the balanced tier, and Luna is the lowest-cost tier.

What is the latest Claude model?

Anthropic's newest major broadly announced releases are Claude Sonnet 5 and Claude Fable 5. Sonnet 5 is the scalable default; Fable 5 is the higher-capability, higher-cost option. Mythos 5 is restricted rather than generally available.

What is the latest Gemini model?

Gemini 3.5 Flash is the released member of Google's Gemini 3.5 family. Google announced that Gemini 3.5 Pro was in development, so availability should be checked before planning production use around it.

How often is this guide updated?

We review first-party release announcements and update the date at the top whenever a significant model launches or its access status changes. Always confirm model IDs, regional availability, and current prices in the provider's documentation before deploying.

Keep comparing

• LLM API pricing — current token prices and source links
• GPT vs Claude vs Gemini — deeper comparison by use case
• Cheapest LLM APIs — options for cost-sensitive workloads