Medium App: 1K requests/day, 3K tokens avg (1K in / 2K out)

Scale App: 5K requests/day, 2K tokens avg (500 in / 1.5K out)

Batch Processing: 10K requests/day, 1K tokens avg (non-urgent)

At every scale, GPT-5 saves 64-70% over Opus 4.7. Even with Opus 4.7's Batch API at 50% off, GPT-5 standard pricing is still 64% cheaper. The premium for Opus 4.7 is steep — $15,000+/mo at scale — and you need a strong justification to absorb it.

Cost per Request by Type

GPT-5 is 60-68% cheaper per request across all workload types. The gap is widest on input-heavy requests (document analysis, RAG) where Opus 4.7's $5 input price dominates the cost. Even on output-heavy content generation, GPT-5 still saves 60%.

The Batch API: Opus 4.7's Best Card

Even at Batch API pricing, Opus 4.7 is still 2x more expensive on input and 25% more on output than GPT-5 standard pricing. GPT-5 doesn't need a Batch API discount to undercut Opus 4.7 — it's cheaper at full price. The Batch API narrows the gap from 4x to 2x on input, but that's still a significant premium.

When Opus 4.7 Justifies the Premium

• 1M context window: When your workloads consistently exceed 272K tokens — analyzing entire codebases, processing long documents, or maintaining massive conversation histories — Opus 4.7's 1M context is a capability GPT-5 simply cannot match
• Coding at the frontier: Opus 4.7 is widely regarded as the best coding model available. For complex multi-file refactoring, architecture planning, and debugging, the quality gap can justify 4x cost — especially for senior engineering tasks where accuracy saves debugging time
• Extended thinking: Opus 4.7's extended thinking mode handles multi-step reasoning chains that require careful planning — useful for mathematical proofs, algorithm design, and complex analysis
• Instruction following precision: For applications with intricate prompt constraints and formatting requirements, Opus 4.7 follows instructions with fewer errors — reducing retry costs that erode GPT-5's price advantage
• AI agent orchestration: Opus 4.7's tool use and function calling capabilities are stronger for complex multi-tool agent workflows where reliability matters more than cost
• Research and analysis: When the cost of being wrong exceeds the cost of the API call — financial analysis, legal document review, medical literature synthesis — Opus 4.7's accuracy premium pays for itself

When GPT-5 Wins: Value and Speed

• Cost-sensitive production: At 60-70% cheaper, GPT-5 delivers premium-tier quality for most workloads at a fraction of the cost. For chatbots, content generation, summarization, and general-purpose AI features, GPT-5 is the pragmatic choice
• High-volume applications: At scale, the savings are massive — $7,875/mo at 5K requests/day. That budget can fund an entire engineering team's AI tooling
• Standard context needs: If your workloads stay under 272K tokens (which covers the vast majority of applications), GPT-5's smaller context window isn't a limitation
• OpenAI ecosystem: Native integration with OpenAI's platform, plugin system, and tooling. If you're already on OpenAI, GPT-5 integrates seamlessly
• Speed: GPT-5 generally offers faster inference times than Opus 4.7 for interactive workloads
• Batch processing: Without needing a Batch API discount, GPT-5 at $1.25/$10 is cheaper than most mid-tier models — making it the default for non-urgent processing

The Decision Framework

Budget Alternatives

Both models are premium-priced. If cost is the primary concern, there are much cheaper options:

Gemini 2.5 Flash-Lite at $0.075/$0.30 with 1M context is 99% cheaper than Opus 4.7. For many production workloads — classification, summarization, simple Q&A — a budget model performs adequately at a fraction of the cost. Start with a budget model, escalate to GPT-5 or Opus 4.7 only when quality requirements demand it.