I keep forgetting which model to reach for and what things cost, so I wrote it down. Numbers are from April 2026 — prices change, so double-check before making decisions.
Models overview
Three main providers plus Cursor, which routes between them automatically for coding.
| Provider | Models | Known for |
|---|---|---|
| Anthropic | Haiku 4.5 · Sonnet 4.6 · Opus 4.6 | Natural writing, long context, coding |
| OpenAI | GPT-4o-mini · GPT-4o · GPT-5.2 · o3 | Broad ecosystem, all-round reliability |
| Gemini 2.5 Flash · 2.5 Pro · 3.1 Pro | Reasoning, long documents, cost efficiency | |
| Cursor | Routes between the above | Coding — picks the model per task |
Use cases
Coding
Claude Opus 4.6 leads on complex, multi-file work — 80.8% on SWE-bench Verified. Gemini 3.1 Pro is close behind (78.8%) at roughly 60% lower cost. Cursor’s auto mode routes mostly to Sonnet 4.6 and Gemini day-to-day, Opus for harder problems.
Reasoning and research
Gemini 3.1 Pro: 94.3% on GPQA Diamond, 77.1% on ARC-AGI-2. OpenAI’s o3 is designed for step-by-step reasoning — good for math and logic.
Writing and long documents
Claude. Most natural prose, up to 128K tokens output per pass. Sonnet 4.6 is the cost-efficient pick here.
High-volume / budget
Gemini 2.5 Flash at $0.075/M input. GPT-4o-mini at $0.15. Both fine for classification, summarization, scale tasks.
General everyday use
GPT-5.2 — widest ecosystem, solid at most things. Gemini 2.5 Pro is close at lower cost.
Vision and image understanding
| Vision task | Best pick |
|---|---|
| Text/OCR from photos | GPT-4o |
| Charts, diagrams, academic figures | Gemini 3.1 Pro |
| Non-English document OCR | Gemini |
| UI screenshots → code | Claude Sonnet/Opus |
| Technical diagram interpretation | Claude or GPT-4o |
| General photo description | GPT-4o or Gemini |
Notes on each:
- GPT-4o — best for text extraction from photos, layout-aware document parsing. Struggles with non-Latin scripts and medical imagery.
- Gemini 3.1 Pro / Flash — 79.0 on MMMU Pro (charts, diagrams, academic figures). Best non-English OCR. Integrates natively with Google Docs/Sheets.
- Claude Sonnet / Opus — best for UI screenshots and generating code from them. Solid on technical diagrams.
API pricing (April 2026)
These are API prices — developer pay-per-token, not subscriptions. ChatGPT Plus / Claude Pro / Gemini Advanced are flat monthly fees regardless of usage.
| Model | Input / 1M tokens | Output / 1M tokens |
|---|---|---|
| Gemini 2.5 Flash | $0.075 | $0.30 |
| GPT-4o-mini | $0.15 | $0.60 |
| Claude Haiku 4.5 | $1.00 | $5.00 |
| GPT-5.2 | $1.75 | $14.00 |
| Gemini 3.1 Pro | $2.00 | $12.00 |
| GPT-4o | $2.50 | $10.00 |
| Claude Sonnet 4.6 | $3.00 | $15.00 |
| Gemini 2.5 Pro | $4.00 | $20.00 |
| Claude Opus 4.6 | $5.00 | $25.00 |
| o3 (reasoning) | $10.00 | — |
Output tokens cost roughly 4–5× more than input. Prompt caching can cut input costs by up to 90%.
Quick pick table
| Task | Reach for |
|---|---|
| Complex multi-file coding | Claude Opus 4.6 or Gemini 3.1 Pro |
| Everyday coding | Claude Sonnet 4.6 (or Cursor auto) |
| Writing and long documents | Claude Sonnet or Opus |
| Step-by-step reasoning | Gemini 3.1 Pro or o3 |
| General-purpose assistant | GPT-5.2 |
| High-volume / budget | Gemini 2.5 Flash or GPT-4o-mini |
| OCR / text from photos | GPT-4o |
| Charts and diagram reading | Gemini 3.1 Pro |
| UI screenshots | Claude Sonnet/Opus |
| Non-English documents | Gemini |
Sources — verify numbers here
- Claude API Pricing — Anthropic Docs
- OpenAI API Pricing
- Gemini API Pricing — Google AI for Developers (official)
- Cursor Models & Pricing — Cursor Docs
- AI Model Benchmarks April 2026 — LM Council
- Best LLM for Coding 2026 — Onyx AI
- Best Vision & Multimodal LLMs 2026 — WhatLLM
- GPT-4o vs Claude vs Gemini: 50 prompts compared — DEV Community