GPT-5.5 Released | OpenAI Doubles API Pricing for Its Smartest Model Yet

OpenAI released GPT-5.5 on April 23, 2026, describing it as their smartest and most intuitive model yet, designed specifically for complex, multi-step tasks across coding, research, and data analysis. The release arrived with a pricing decision that immediately sparked debate: API costs doubled compared to GPT-5.4, making GPT-5.5 the most expensive frontier model at standard tier.

The model is included in ChatGPT Plus at $20/month with no additional charge, which means consumer users get access without a cost increase. The price jump is felt entirely by developers and enterprises using the API, a bifurcation that reflects OpenAI's different economics on each side of the market.

API Pricing | A Full Breakdown

At launch, GPT-5.5 is priced at $5 per million input tokens and $30 per million output tokens at the standard tier, a 2x jump from GPT-5.4's $2.50/$15. GPT-5.5 Pro, the extended capability tier, comes in at $30/$180 per million tokens, a level that puts it in a different budget category than anything previously available from OpenAI or its competitors.

For context: at these prices, a production application making moderate use of the API (say, 10 million output tokens per month) would spend $300/month on GPT-5.5 standard versus $150 on GPT-5.4 or roughly $250 on Claude Opus 4.8. At GPT-5.5 Pro pricing, the same workload costs $1,800/month, a number that limits adoption to large-scale production applications where the capability delta justifies the cost.

Agentic Capability: The Core Bet

GPT-5.5 is described by OpenAI as built specifically for agentic workflows, multi-step, tool-using operations where the model plans and executes over long time horizons. The improvements over GPT-5.4 are most pronounced in three areas: error recovery (the model identifies and corrects its own mistakes more reliably), tool use efficiency (fewer redundant API calls in multi-step tasks), and extended context handling (better coherence across long conversations and documents).

For developers building autonomous agents, these are the improvements that matter. GPT-5.4 was already capable, but it could lose coherence in very long agentic tasks and sometimes made redundant tool calls that increased both latency and cost. GPT-5.5 addresses these failure modes more directly than any previous OpenAI release.

Benchmarks vs Claude Opus 4.8 and Gemini 3.1

On published benchmarks, GPT-5.5 leads on several agentic coding tasks and scores competitively on multi-step reasoning. On SWE-bench Verified, GPT-5.5 scores below Claude Opus 4.8 (88.6%) but ahead of Gemini 3.1 Pro on the same version of the benchmark. On GPQA Diamond (a graduate-level reasoning benchmark), the three models are within a few percentage points of each other, close enough that task-specific testing matters more than headline numbers.

Speed is where GPT-5.5 has a clearer edge: OpenAI's inference infrastructure remains the fastest in the market at standard tier, with median time-to-first-token consistently below Claude and Gemini for equivalent prompt lengths.

How It Affects the Rankly Leaderboard

On Rankly, the pricing increase has a direct negative impact on GPT-5.5's B5 Accessibility score, our f(A) dimension that tracks price, latency, and availability. A 2x price increase on the most-used API tier is a meaningful penalty in the f(A) calculation, even accounting for the fast inference speeds.

On the other side, B1 Intelligence (driven by benchmark results) moves higher following the agentic capability improvements. The net effect is that GPT-5.5's RQ score reflects a model that is stronger on raw capability but meaningfully less accessible than its predecessor, a trade-off that shows up clearly in the leaderboard rankings.

Verdict

GPT-5.5 is a legitimate capability upgrade over GPT-5.4, particularly for agentic use cases. The question is whether the capability delta justifies a 2x price increase when Claude Opus 4.8, released five weeks later, offers comparable or better benchmark performance at $5/$25 (vs $5/$30 for GPT-5.5).

For teams already deeply integrated with OpenAI's tooling, APIs, and fine-tuning infrastructure, GPT-5.5 is worth the upgrade. For teams evaluating models fresh, the pricing decision makes it harder to recommend over Claude Opus 4.8 unless you specifically need OpenAI's speed or ecosystem integrations.