Anthropic Quietly Upgrades Free Claude AI to Outperform ChatGPT

Anthropic has made its free offering on claude.ai significantly more powerful than OpenAI's ChatGPT free tier. The company quietly switched the default model for free users to its new Claude Sonnet 5.5, which outperforms the GPT-6 Luna model currently powering ChatGPT's free version.

This move happened alongside the model's public release. Anthropic positions Sonnet 5.5 as a direct successor to Sonnet 5. The company claims it runs 30% or more faster and costs up to 30% less for most workloads. Critically, the API pricing remains identical to the previous Sonnet model. You get better performance for the same money.

Benchmarks appear to confirm the claims. Early testing shows Sonnet 5.5 beating its predecessor on nearly every standard test. On certain coding tasks, it performs close to the more expensive Opus 5.5 model. This is a notable achievement for a model in the Sonnet, or mid-tier, category.

The cost efficiency is a key development for builders. If the model completes tasks faster, your latency drops. If it costs the same per token but requires fewer tokens for the same result, your overall expense falls. This combination is rare in the crowded LLM market.

Simon Willison's blog highlights an interesting bug. Both Sonnet 5.5 and Opus 5.5 struggle with a specific, complex prompt: generating an SVG of a pelican riding a bicycle using "max" thinking effort. The model can exhaust its 128,000 token thinking budget without producing output, costing $1.28 per failed attempt. This reveals a current limit in handling very long, chain-of-thought reasoning for visual generation.

Using a lower, "xhigh" thinking setting works. It produced a valid SVG in 41 seconds for a cost of 5.74 cents. The difference shows how parameter choices, like thinking effort, have direct cost and reliability implications for developers.

The real competition is now in the free consumer space. OpenAI's GPT-6 Luna was the clear leader for free users. Anthropic, with Sonnet 5.5, has taken that lead. A simple test, asking for a 3D WebGL pelican on a bicycle, returned a solid functional page from the free claude.ai.

For developers, this signals that high-quality models are becoming table stakes. The "good enough" bar for free tiers is rising rapidly. It also means Anthropic is investing heavily in its user base, hoping to convert free users into API customers.

The community context matters. This release follows closely after the debut of Claude Opus 5.5, GPT-6 Sol, and GPT-6 Luna. A recent price war between Anthropic and OpenAI is already underway. Sonnet 5.5 intensifies that pressure.

One model is still missing from Anthropic's lineup: Haiku 5.5. The company announced it will arrive "in the coming weeks." The big question is its price point. If Haiku 5.5 is priced to compete with GPT-6 Luna, it could trigger another significant shift in the API cost landscape.

The immediate takeaway is the free tier change. It forces other providers to respond. How long can OpenAI sustain a free model that is now demonstrably weaker? The competition will likely force more performance improvements at lower price points for everyone building with AI.

Source: Claude Sonnet 5.5