GPT-5.5 Is Here: OpenAI's Smartest Model, Available Now on Smart AIPI

OpenAI just released GPT-5.5, their smartest and most intuitive model yet. We added it to Smart AIPI the same day. Use model ID gpt-5.5 in any chat completion or responses API call and you're using it — no config changes, no redeployment, same API key.

Below: the benchmark numbers that matter, how it compares to GPT-5.4, Claude Opus 4.7, and Gemini 3.1 Pro, and why Smart AIPI is the cheapest place to use it.

The Benchmarks: GPT-5.5 Sets New State of the Art

GPT-5.5 improves on GPT-5.4 across every major benchmark. The gains are especially strong in agentic coding, computer use, knowledge work, and early scientific research — areas where progress depends on reasoning across context and taking action over time.

The standout numbers: 82.7% on Terminal-Bench 2.0 (a 7.6-point jump over GPT-5.4 and 13+ points ahead of Claude Opus 4.7), 84.9% on GDPval knowledge work, 85.0% on ARC-AGI-2 (up from 73.3%), and 35.4% on FrontierMath Tier 4 — the hardest math benchmark in production — up from 27.1% and nearly double Claude Opus 4.7's 22.9%.

What Actually Changed

Three things developers will notice right away:

• Stronger agentic coding. GPT-5.5 carries more of the work itself. It holds context across large systems, reasons through ambiguous failures, checks assumptions with tools, and carries changes through the surrounding codebase. Early NVIDIA engineers said "losing access to GPT-5.5 feels like I've had a limb amputated." On Expert-SWE (long-horizon tasks with a median 20-hour expert completion time), GPT-5.5 hits 73.1% vs GPT-5.4's 68.5%.
• Fewer tokens for the same work. GPT-5.5 is more token-efficient than GPT-5.4 on Codex tasks. Terminal-Bench 2.0 at 82.7% uses fewer tokens per task. Expert-SWE at 73.1% also uses fewer tokens. That means the raw per-token price is only half the story — your effective cost per finished task is lower still.
• Same latency. Larger, more capable models are typically slower to serve, but GPT-5.5 matches GPT-5.4 per-token latency in real-world serving. You get a smarter model without paying for it in wall-clock time.

Beyond raw benchmarks, senior engineers testing the model have reported a step change: GPT-5.5 catches issues in advance, predicts testing and review needs without explicit prompting, and resolves complex merges in a single shot. One CEO described it as "the first coding model I've used that has serious conceptual clarity."

Smart AIPI Is the Cheapest Way to Use GPT-5.5

OpenAI's direct API pricing for GPT-5.5 is $5 per 1M input tokens and $30 per 1M output tokens. That's a 2x jump over GPT-5.4's direct pricing, reflecting the capability leap.

Through Smart AIPI, you pay:

A typical Codex-style agentic workload with 50K input tokens, 20K cached prefix hits, and 8K output tokens costs $0.32 through Smart AIPI versus $1.28 direct from OpenAI. Across a day of 500 such requests, that's $160 vs $640 — a $480 daily difference for the exact same model running on the exact same OpenAI infrastructure.

No other provider offers GPT-5.5 at this price. OpenRouter passes through OpenAI direct pricing plus a margin. Google Vertex and Azure OpenAI charge the full direct rate. Smart AIPI aggregates ChatGPT Business subscription capacity and passes the savings to you.

How to Use GPT-5.5 on Smart AIPI

It's already live. Just set your model parameter:

Chat Completions API

curl https://api.smartaipi.com/v1/chat/completions \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-5.5",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

Responses API

curl https://api.smartaipi.com/v1/responses \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-5.5",
    "input": "Refactor this function for clarity and add tests"
  }'

With High Reasoning Effort

curl https://api.smartaipi.com/v1/responses \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-5.5",
    "input": "Design a caching strategy for a high-read API...",
    "reasoning": {"effort": "high", "summary": "auto"}
  }'

Codex CLI

Update your ~/.codex/config.toml:

model = "gpt-5.5"

Cursor, Cline, Roo Code, OpenCode, Aider

Point the tool at Smart AIPI's base URL and set the model to gpt-5.5:

OPENAI_BASE_URL=https://api.smartaipi.com/v1
OPENAI_API_KEY=sk-your-smart-aipi-key
OPENAI_MODEL=gpt-5.5

Context Window and Advanced Features

• Context window: 400K tokens in Codex mode, 1M tokens via the Responses API.
• Reasoning effort: low, medium, high. For agentic and long-horizon work, high effort is recommended — it's where GPT-5.5 pulls furthest ahead of GPT-5.4.
• Streaming: supported on both chat completions and responses APIs. Use "stream": true.
• Prompt caching: Smart AIPI preserves prompt cache hits upstream. Typical sustained workloads hit 85–95% cache rates on gpt-5.5, so your effective input cost is closer to $0.125 per 1M than $1.25.

What About GPT-5.5 Mini, Nano, and Pro?

OpenAI hasn't released gpt-5.5-mini, gpt-5.5-nano, or gpt-5.5-pro on the Codex backend yet. When they go live we'll add them the same day and publish a follow-up. For now, only the base gpt-5.5 model is available.

If you're using one of the older gpt-5-nano or gpt-5.4-nano aliases in Roo Code, Cline, or similar agent frameworks, Smart AIPI transparently routes those to gpt-5.4-mini for reliability — no action needed on your end.

Get Started

1. Sign up at smartaipi.com/signup (free credits, no credit card)
2. Create an API key in the dashboard
3. Set your base URL to https://api.smartaipi.com/v1
4. Use model gpt-5.5

Same API, same OpenAI-compatible shape, same tools — now with the smartest model OpenAI has ever shipped, at 75% off.