Anthropic may release Claude Sonnet 5.5 within days

September looks like an exciting month for AI, with major new models dropping frequently. And according to recent leaks, Anthropic is nearing the launch of its next significant update. Rumor has it that Claude Sonnet 5.5 is almost ready for release.

Trusted leaker "Lyra" on X reports that Anthropic partners have received a second checkpoint, which seems to perform even better than the first stealth build. They also mentioned that it could arrive as soon as early next week, possibly before Monday morning ends.

This release would follow Anthropic's Claude Opus 5.5 model, which is already winning over the AI community with its impressive performance on various tasks. Early users like its speed and coding skills, and real-world tests show it can deliver better results than OpenAI's Astra while costing less. In the past, users complained about stricter usage limits on Claude during busy times, so a strong new model could help ease those frustrations.

Sonnet has been popular because it offers a strong mix of price and performance, so most people will want to see how Sonnet 5.5 compares to GPT-6. OpenAI recently released GPT-6 Sol and Luna to give developers more affordable choices, but Anthropic looks ready to grab attention just before OpenAI DevDay, where the company is expected to unveil "o," its always-on agent to compete with SpaceXAI's Grok Bot and Muse from Meta.

One thing to note that this also comes at a time when OpenAI is has been dealing with safety issues and problems with its sandbox in the background.

In a recent alignment blog post about an agent bypassing network restrictions to access outside services, OpenAI said, "All training, evaluation, and inference with tool-use (defined broadly) of our most capable models remain paused." This pause is a major setback for their latest rollout.

Meanwhile, Anthropic is clearly trying to maintain its lead in the AI race with its increasingly powerful models. With Opus 5.5, and now Sonnet 5.5 on the horizon, Anthropic is setting the benchmark for AI performance yet again. In an earlier post, Lyra reported that Sonnet 5.5 pricing at the time was: Input: $2, Output: $10, and Cache reads: $0.20/M. This could easily change when the model is released publicly, so take that with a grain of salt.

AI tester Chetaslua also shared an alleged video comparison of Sonnet 5.5 (currently appears to being A/B tested randomly on Claude) and GPT-6 Sol with the same prompt:

That said, OpenAI is not the only company under pressure. Google may soon join the competition with its Gemini 4 Pro model.

The company has already shown strong browser automation results with its Flash tier, but people have been waiting months for a true flagship upgrade. Early checkpoint test results look promising, especially with cleaner frontend designs that avoid the robotic style of older models.

Google told The Information that Gemini 4 should be available well before the end of the year, so it will likely arrive in the next few weeks.

With Sonnet 5.5 about to launch, OpenAI working on its safety sandbox, and Google preparing Gemini 4 Pro, the next few weeks will be busy.