- OpenAI released GPT-5.5, codenamed ‘Spud,’ this week, reporting leading benchmark scores across reasoning, agentic tasks, computer use, and coding for publicly available models.
- The model is priced at $5 per million input tokens and $30 per million output tokens, which OpenAI describes as “half the cost of competitive frontier coding models.”
- GPT-5.5 maintains GPT-5.4’s inference speed while adding efficiency gains, with OpenAI stating it used Codex and GPT-5.5 itself to rewrite its GPU infrastructure code.
- Anthropic separately published a post-mortem this week attributing Claude Code quality complaints to three distinct bugs and reset usage limits for affected subscribers.
What Happened
OpenAI this week released GPT-5.5 — internally codenamed ‘Spud’ — describing it as “a new class of intelligence” in a positioning statement reported by The Rundown AI on April 25, 2026. The company reported that GPT-5.5 posts leading benchmark scores across reasoning, agentic, computer use, and coding evaluations for publicly available models, with several results described as comparable to Anthropic’s Claude Mythos. The release comes as Anthropic faces user-reported quality degradation and rate-limiting complaints across its Claude lineup.
Why It Matters
The launch marks a competitive shift following a period in which Anthropic’s Claude models were broadly cited at the top of public AI benchmark rankings. OpenAI CEO Sam Altman has publicly emphasized the company’s accelerated shipping cadence in 2026, and GPT-5.5 lands at a moment when Anthropic acknowledged this week that three separate bugs had degraded Claude Code quality and forced a usage-limit reset for subscribers. The convergence of GPT-5.5’s release and Anthropic’s technical difficulties sharpens the competitive contrast in developer-facing AI tooling.
Technical Details
According to The Rundown AI’s reporting, GPT-5.5 sets new highs across reasoning, agentic, computer use, and coding tests among public models, with benchmark scores in several categories described as comparable to Claude Mythos. OpenAI stated it used both Codex and GPT-5.5 to rewrite its GPU infrastructure code, presenting the process as a practical demonstration of the model’s agentic and coding capabilities applied internally. The API is priced at $5 per million input tokens and $30 per million output tokens — which OpenAI claims is “half the cost of competitive frontier coding models.” The model is rolling out across ChatGPT subscription plans and inside Codex, with Thinking and Pro variants included in the initial release wave.
Who’s Affected
Developers building agentic and coding applications on the OpenAI API are the immediate audience, given the model’s stated focus on those task categories. Enterprise customers evaluating frontier models will be able to benchmark GPT-5.5’s price-to-performance ratio directly against Claude Mythos and other competing offerings. OpenAI indicated GPT-5.5 will carry what it characterized as “generous usage” across ChatGPT subscription tiers, which could influence individual developer and researcher adoption decisions.
What’s Next
OpenAI has not disclosed a timeline for further GPT-5.x releases beyond the current GPT-5.5 rollout. Anthropic’s published post-mortem addresses the immediate Claude Code bugs and limit resets but does not specify a model update schedule. Independent third-party benchmark evaluations comparing GPT-5.5 against Claude Mythos and other frontier models are expected as broader API access is established.
