OpenAI’s April 23, 2026 release of GPT-5.5 doubled the per-token price compared to GPT-5.4: input tokens now cost $5.00/M, output tokens $30.00/M. Pro API pricing remains at $30/$180. Knowing the detailed pricing surfaces—API, Batch, Flex, Priority, and Codex limits—is essential to avoid surprise bills and optimize your workload before upgrading to GPT-5.5.
For a high-level overview, see What is GPT-5.5. For API integration steps, check How to use the GPT-5.5 API.
TL;DR
| Surface | Input / M | Output / M |
|---|---|---|
| GPT-5.5 standard API | $5.00 | $30.00 |
| GPT-5.5 Pro API | $30.00 | $180.00 |
| GPT-5.5 Batch (50% off) | $2.50 | $15.00 |
| GPT-5.5 Flex (50% off) | $2.50 | $15.00 |
| GPT-5.5 Priority (2.5×) | $12.50 | $75.00 |
| GPT-5.4 standard API | $2.50 | $15.00 |
| GPT-5.4-mini API | $0.25 | $2.00 |
GPT-5.5 costs 2× GPT-5.4 per token, but OpenAI claims ~20% higher net intelligence when factoring in token efficiency.
The Headline Numbers
OpenAI’s API pricing page lists current rates:
- GPT-5.5: $5.00/M input, $30.00/M output tokens.
- GPT-5.5 Pro: $30.00/M input, $180.00/M output tokens.
- Context window: 1M tokens for both variants. Reasoning tokens count toward both context and output billing.
Batch, Flex, and Priority
OpenAI provides alternative pricing tiers for specific workload needs:
Batch API
Use the Batch endpoint for queued requests at 50% standard pricing. Turnaround is <24 hours. Ideal for:
- Overnight dataset evaluations.
- Historical re-processing.
- Any workflow with latency tolerance measured in hours.
Batch pricing: GPT-5.5 at $2.50 / $15.00 per million tokens—same as GPT-5.4 standard. Use Batch for offline workloads to avoid the price hike.
Flex Processing
Flex offers 50% off standard rates with variable latency, from seconds to several minutes depending on demand. Choose Flex for unpredictable latency tolerance with Batch-level savings and near-synchronous responses.
Priority Processing
Priority tier charges 2.5× standard rates ($12.50/$75.00 per million tokens for GPT-5.5) in exchange for faster throughput, higher rate limits, and minimal queue time. Reserve for latency-critical, user-facing production scenarios.
Thinking Mode Cost Math
reasoning.effort increases the number of tokens used per request, not the per-token price. Adjust your math based on effort:
| Effort | Output-token multiplier | When to use |
|---|---|---|
low |
1× | Routine calls |
medium |
1.3–2× | Multi-step coding, structured generation |
high |
2–4× | Deep research, correctness-critical review |
xhigh |
3–8× | Agent loops, dense planning |
A single xhigh call on a long prompt may use 20K reasoning tokens—$0.60 just for reasoning at standard output rates. Budget by workload, not per request.
Codex Pricing
Codex access ties to your ChatGPT plan, not token billing. As of April 2026:
| Plan | Codex access | GPT-5.5 | Notes |
|---|---|---|---|
| Free | Yes (limited time) | Yes | Tight weekly caps |
| Go | Yes (limited time) | Yes | 2× Free caps |
| Plus ($20/mo) | Yes | Yes | Standard caps |
| Pro ($200/mo) | Yes | Yes + Thinking + Pro (in ChatGPT) | Highest per-user caps |
| Business | Yes | Yes | Seat-based |
| Enterprise/Edu | Yes | Yes | Contract-based |
For CLI-based terminal workflows, Plus or Pro is the most cost-effective way to access GPT-5.5 after a few hundred thousand tokens/day. See the free path guide for no-cost entry.
Comparison: GPT-5.5 vs Other Models
Choose based on your workload’s output and risk profile:
| Model | Input / M | Output / M | Cost per 1K output tokens |
|---|---|---|---|
| GPT-5.4-mini | $0.25 | $2.00 | $0.0020 |
| GPT-5.4 | $2.50 | $15.00 | $0.0150 |
| GPT-5.5 | $5.00 | $30.00 | $0.0300 |
| GPT-5.5 Pro | $30.00 | $180.00 | $0.1800 |
- High-volume, low-risk output (classification, summarization): GPT-5.4-mini
- General workloads where GPT-5.4 is sufficient: GPT-5.4
- Complex coding, agentic tasks, research: GPT-5.5
- Correctness-critical output: GPT-5.5 Pro
Worked Example: Coding Agent Cost per Task
Typical agentic coding session on GPT-5.5 (reasoning.effort: "medium"):
- Input tokens: ~15,000
- Output tokens: ~3,000
- Reasoning tokens: ~6,000 (medium effort)
Cost per task (standard pricing):
- Input: 15K × $5.00/M = $0.075
- Output: (3K + 6K) × $30.00/M = $0.27
- Total: $0.345 per coding task
Same workload on GPT-5.4:
- Input: 15K × $2.50/M = $0.0375
- Output: 9K × $15.00/M = $0.135
- Total: $0.1725 per task
GPT-5.5 is 2× the cost per task at equal reasoning. If GPT-5.5 closes more tasks successfully (higher quality), the upgrade may pay for itself by reducing retries.
Day 1 Cost Controls to Implement
To manage GPT-5.5 costs, build these controls into your stack:
-
Enforce
max_output_tokenscaps: Default to 2,000 unless longer output is necessary. - Strict JSON schemas: Prevents malformed output and expensive retries.
- Route by difficulty: Use GPT-5.4-mini for easy requests; escalate hard ones to GPT-5.5.
- Use Batch for offline jobs: Evaluations, reports, etc. get 50% off.
-
Monitor
usage.reasoning_tokens: High-effort reasoning tokens are the most common source of bill spikes.
Per-Plan Monthly Cost Estimate
If you’re choosing a ChatGPT plan for GPT-5.5, see the breakdown:
| Plan | Monthly Price | Best Fit |
|---|---|---|
| Free | $0 | Trying GPT-5.5 via Codex |
| Go | $4/mo | Students, light users |
| Plus | $20/mo | Devs using Codex + ChatGPT daily |
| Pro | $200/mo | Power users needing Thinking/Pro modes |
| Business | $25/seat/mo | Teams needing shared workspaces |
| Enterprise/Edu | Custom | Contracted, SLA-based use |
If your API usage exceeds ~4M output tokens/month, Pro plus Codex CLI is usually cheaper—as long as your context fits in the 400K CLI window.
Price Change Signals to Watch
Track these if budgeting long-term:
- GPT-5.5 API general availability: Pricing may drop in response to competition (Claude Mythos, Gemini 3.5, open weights).
- Pro model democratization: OpenAI has historically lowered Pro-tier prices 3–6 months post-launch. Don’t assume $30/$180 is permanent.
FAQ
Does caching reduce input cost?
Yes. Cached input tokens are billed at a fraction of standard rate. Check the OpenAI pricing page for details. Cache reusable system prompts and repo contexts.
Is there a volume discount?
Not officially. Enterprise contracts may have custom rates. For sustained, large-scale use, talk to sales.
Does thinking mode cost extra?
No. It increases token usage, not the per-token rate.
Is Codex CLI usage billed separately?
Only if you sign in with an API key. ChatGPT sign-ins use the plan fee; API keys use usage-based billing.
What’s the cheapest way to try GPT-5.5?
Free or Go plan plus Codex CLI. See our free path guide for details.
Top comments (0)