AI Coding Assistants Pricing Comparison: OpenAI Codex, Claude API & GitHub Copilot
Comprehensive pricing comparison of three major AI coding platforms based on official sources: OpenAI Codex, Anthropic Claude API, and GitHub Copilot. Includes individual plans, enterprise options, and token-based billing models.
Executive Summary
Three major players dominate the AI coding assistant landscape: OpenAI Codex, Anthropic Claude API, and GitHub Copilot. Each employs distinct pricing models reflecting their target audiences and feature sets. Codex uses token-based credits with hybrid subscription tiers; Claude API uses per-million-token (MTok) pricing with feature-specific multipliers; GitHub Copilot recently transitioned to usage-based premium requests from subscription-based models. This comparison draws exclusively from official pricing pages as of April 2026.
1. OpenAI Codex: Hybrid Subscription + Token Model
Overview
OpenAI's Codex pricing emphasizes accessibility for individuals while scaling through a hybrid system combining subscription tiers with optional credit purchases.
Official Source: https://developers.openai.com/codex/pricing
Consumer Plans (Individual Users)
| Plan | Monthly Price | Codex Usage Limits | Key Features |
|---|---|---|---|
| Free | $0 | Explore capabilities | Basic exploration |
| Go | ~$20/month | Lightweight tasks | CLI, IDE extension, iOS |
| Plus | Bundled with ChatGPT Plus | 15-80 messages/5h (depends on model) | Web, CLI, IDE, iOS; Cloud features; Latest models |
| Pro | $100 or $200/month | 10x-25x Plus limits | Higher rate limits; Fast Codex models; 2x promo through May 31, 2026 |
Token-Based Credits (New as of April 2, 2026)
OpenAI transitioned to token-based credit pricing effective April 2, 2026, replacing per-message estimates. Applies to new and existing Business customers, and new Enterprise customers.
Credits per 1M Tokens:
| Model | Input | Cached Input | Output |
|---|---|---|---|
| GPT-5.5 | 125 | 12.50 | 750 |
| GPT-5.4 | 62.50 | 6.25 | 375 |
| GPT-5.4-mini | 18.75 | 1.875 | 113 |
| GPT-5.3-Codex | 43.75 | 4.375 | 350 |
| GPT-5.2 | 43.75 | 4.375 | 350 |
Pricing Implications:
- Fast mode increases credit consumption (varies by model)
- Cached input tokens cost 10% of standard input rate
- Image generation uses 3-5x faster consumption
API Key Pricing
Pay-per-token model for CI/CD and automated environments:
- No cloud features (GitHub code review, Slack integration)
- Delayed access to latest models
- Standard API pricing applies
Business/Enterprise
- ChatGPT Business: $20/seat/year (down from $25)
- Enterprise: Custom pricing; no fixed rate limits
- $500 credit promotion for eligible existing ChatGPT Business workspaces
2. Anthropic Claude API: Per-Million-Token Pricing with Caching
Overview
Claude API employs transparent per-token pricing with specialized multipliers for caching, batch processing, and premium features. Pricing varies significantly by model and context window.
Official Source: https://platform.claude.com/docs/en/about-claude/pricing
Standard Model Pricing (Per Million Tokens)
| Model | Input | Output | Cache Read (5m) | Cache Write (1h) |
|---|---|---|---|---|
| Claude Opus 4.7 | $5 | $25 | $0.50 | $5.00-$10.00 |
| Claude Opus 4.6 | $5 | $25 | $0.50 | $5.00-$10.00 |
| Claude Sonnet 4.6 | $3 | $15 | $0.30 | $3.00-$6.00 |
| Claude Haiku 4.5 | $1 | $5 | $0.10 | $1.00-$2.00 |
| Claude Haiku 3.5 | $0.80 | $4 | $0.08 | $0.80-$1.60 |
Feature-Specific Pricing
Prompt Caching Multipliers:
- 5-minute cache write: 1.25x base input
- 1-hour cache write: 2x base input
- Cache hit (read): 0.1x base input (10% of standard)
- Breaks even after 1 cache read (5m) or 2 reads (1h)
Batch Processing (50% Discount):
| Model | Batch Input | Batch Output |
|---|---|---|
| Claude Opus 4.7 | $2.50 | $12.50 |
| Claude Sonnet 4.6 | $1.50 | $7.50 |
| Claude Haiku 4.5 | $0.50 | $2.50 |
Fast Mode Pricing (Opus 4.6 only, Beta):
- Input: $30/MTok (6x standard)
- Output: $150/MTok (6x standard)
- Not available with Batch API
Data Residency (US-only):
- 1.1x multiplier on all token categories
- Applies to Claude 4.6 and newer models only
- Third-party platforms (AWS Bedrock, Google Vertex) have separate regional pricing
Long Context (No Premium)
Full 1M token context window at standard pricing (Claude Opus 4.7, Sonnet 4.6):
- 900k-token request = 9k-token request (same per-token rate)
- Caching and batch discounts apply at standard rates
Tool-Specific Pricing
- Bash tool: +245 input tokens per call
- Web search: $10 per 1,000 searches (+ standard token costs)
- Web fetch: No additional cost (token-based only)
- Code execution: Free with web search/fetch; otherwise $0.05/hour/container (1,550 free hours/month)
- Computer use: 466-499 system prompt tokens + 735 tool definition tokens
Claude Managed Agents Billing
Dual billing structure:
- Standard token pricing (with caching multipliers)
- Session runtime: $0.08 per session-hour (running status only)
3. GitHub Copilot: Usage-Based Premium Requests
Overview
GitHub Copilot recently transitioned (June 1, 2026) from request-based to usage-based billing. Individual plans replaced with Free/Pro/Pro+ tiers; Business and Enterprise use premium request (PRU) multipliers.
Official Source: https://github.com/features/copilot/plans
Individual Plans (Pricing effective 2026)
| Plan | Monthly Price | Premium Requests | Chat/Agent Requests |
|---|---|---|---|
| Free | $0 | 50/month | 50 agent mode + 2,000 completions/month |
| Pro | $10 | 300/month | Unlimited with models |
| Pro+ | $39 | 1,500/month | Unlimited with all models incl. Claude Opus 4.7 |
Additional Premium Requests: $0.04 per request
Business & Enterprise (Custom Licensing)
- GitHub Copilot Business: Org seat-based licensing with policy management
- GitHub Copilot Enterprise: Business + codebase indexing + custom fine-tuned models + github.com integration
- Custom pricing: Contact sales for volume deals
Premium Request Multipliers (for Business/Enterprise)
Premium requests consumed at different rates depending on feature:
- Agent mode, code review, coding agent, Copilot CLI: Variable by model
- Chat interactions: Variable by model selected
- Codex/Claude models: Different multiplier than GPT-5 models
Standard multiplier: 1 PRU = $0.01 USD equivalent
Available Models Across All Plans
All three tiers (Free, Pro, Pro+) include access to:
- Anthropic Claude: Haiku 4.5, Sonnet 4.6, Opus 4.7 (Pro+ only for Opus)
- OpenAI: GPT-5 variants, GPT-5.5, GPT-5.4, GPT-5.3-Codex
- Google: Gemini 2.5 Pro, Gemini 3 Flash
- xAI: Grok Code Fast 1
Feature Availability
| Feature | Free | Pro | Pro+ |
|---|---|---|---|
| Inline suggestions | 2,000/mo | Unlimited | Unlimited |
| Agent mode | 50/mo | Unlimited | Unlimited |
| Code review | ❌ | ✅ | ✅ |
| Cloud agent | ❌ | ✅ | ✅ |
| Copilot CLI | ❌ | ✅ | ✅ |
| All models | Limited | Yes | Yes (incl. Claude Opus 4.7) |
| GitHub Spark | ❌ | ❌ | ✅ |
Billing Transition (June 1, 2026)
- Annual Pro/Pro+ subscribers: Remain on existing plans until expiration; model multipliers increase June 1
- At expiration: Transition to Free with option to upgrade to monthly plan
- Monthly subscribers: Already on usage-based billing
4. Comparative Cost Analysis
Scenario 1: Small Team Development (1M tokens/month)
Assumptions:
- 1 million input tokens + 500k output tokens monthly
- Single developer, standard models (no premium features)
| Platform | Model | Monthly Cost |
|---|---|---|
| Codex | GPT-5.4 | ~$31 (Plus tier) |
| Claude API | Sonnet 4.6 | $4.50 (input: $3, output: $7.50) |
| GitHub Copilot | Free tier | $0 (50 req/mo limit) → Pro: $10 |
Winner: Claude API (lowest per-token; Copilot free for limited use)
Scenario 2: Enterprise Deployment (100M tokens/month, with caching)
Assumptions:
- 100M input + 50M output tokens
- 40% cache hit rate on input (40M cached inputs)
- Batch processing for 20% of requests
| Platform | Model | Calculation | Monthly Cost |
|---|---|---|---|
| Codex | GPT-5.4 (Business) | Token credits at 62.5/1M input, 375/1M output | ~$8,750 |
| Claude API | Sonnet 4.6 with batch + cache | (60M × $3) + (40M × $0.30 cache) + (50M × $7.50 batch) | $370 |
| GitHub Copilot | Enterprise (custom) | Estimated 1,500k PRUs @ ~$0.04+ | ~$6,000-10,000 |
Winner: Claude API (caching + batch = 50% savings; 20-25x cheaper than Codex)
Scenario 3: High-Frequency Chat (Agent Mode, 1,000 daily requests)
Assumptions:
- 30,000 premium requests/month
- Mix of GPT-5.4 and Claude Sonnet 4.6
| Platform | Model | Monthly Cost |
|---|---|---|
| Codex | Pro $200 | ~$200 (20x Plus allowance; overage ~$25/1M beyond) |
| Claude API | Sonnet API calls | ~$180 (30k calls × avg 5k tokens @ $3 input, $15 output) |
| GitHub Copilot | Pro: 300 requests | Pro: $10 + $1,080 overage ($0.04 × 27,000 extra) |
Winner: Codex or Claude (subscription-based models better for high volume)
5. Key Pricing Differentiators
OpenAI Codex
Strengths:
- Hybrid model reduces friction for occasional users (subscription + pay-as-you-go)
- Fast Codex model (GPT-5.3-Codex-Spark) optimized for daily tasks
- $500 business credit promotion lowers entry cost
Weaknesses:
- Token-based credits less transparent than pure per-token pricing
- Fast mode unavailable on API tier
- Higher per-token rates than Claude for equivalent models
Claude API
Strengths:
- Transparent per-token pricing (easiest to predict costs)
- Aggressive caching discounts (10% of standard on hits)
- Batch API 50% discount for asynchronous work
- Long context (1M tokens) included at no premium
- Lowest absolute prices for Haiku/Sonnet models
Weaknesses:
- Fast mode limited to Opus 4.6 (expensive at 6x)
- Web search $10 per 1k searches adds up at scale
- No subscription option (all per-use)
GitHub Copilot
Strengths:
- Free tier genuinely useful (50 PRU + 2,000 completions/month)
- Single unified platform (IDE + github.com native)
- Wide model access (OpenAI, Claude, Google, xAI)
- Enterprise features (custom models, IP indemnity)
Weaknesses:
- Premium request model opaque (varies by feature/model)
- Transition away from fixed subscriptions increases unpredictability
- Pro tier ($10/mo) requires additional PRU purchases for heavy use
- Enterprise pricing requires sales contact
6. Best-Fit Recommendations
Choose Claude API if:
- Building production AI features with cost optimization
- Using large documents/long context (1M tokens)
- Leveraging batch processing for non-real-time tasks
- Need predictable, lowest-cost per-token rates
Choose OpenAI Codex if:
- Integrating directly into ChatGPT workflows
- Need fast inference (GPT-5.3-Codex-Spark)
- Prefer hybrid subscription + pay-as-you-go model
- Business promotion credits available
Choose GitHub Copilot if:
- Individual developers wanting single platform (IDE + web)
- Teams needing unified license management
- Prefer access to multiple model vendors
- Enterprise governance/custom models important
7. Pricing Trends & Future Outlook
Observed trends (April 2026):
- Caching as competitive advantage: Claude's aggressive prompt caching (0.1x reads) incentivizes stateful applications
- Batch processing premiumization: Both Claude (50% discount) and Codex (credits model) encourage off-peak usage
- Model diversity: GitHub consolidating multiple APIs (Claude, GPT-5, Gemini) into single platform
- Enterprise custom pricing: All three providers offer custom rates for high-volume customers
- Shift away from fixed subscriptions: Copilot moving to usage-based; Codex already mixed; Claude API pure usage
Cost optimization strategies:
- Prefer Claude for large documents + caching
- Use Batch API for asynchronous workloads (50% savings)
- Leverage free tiers (Copilot Free, Codex Go) for prototyping
- Monitor cached input hit rates (breakeven: 1 read for 5m, 2 reads for 1h)
8. References & Sources
-
OpenAI Codex Pricing (April 2026)
https://developers.openai.com/codex/pricing- Token-based credit model effective April 2, 2026
- ChatGPT Business pricing: $20/seat/year
-
Anthropic Claude API Pricing (April 2026)
https://platform.claude.com/docs/en/about-claude/pricing- Per-million-token pricing with feature multipliers
- Batch processing (50% discount), caching (0.1x reads), fast mode (6x)
-
GitHub Copilot Plans & Pricing (April 2026)
https://github.com/features/copilot/plans- Free: $0 (50 PRU + 2k completions/month)
- Pro: $10/month (300 PRU)
- Pro+: $39/month (1,500 PRU)
- Enterprise: Custom pricing
Article compiled: April 29, 2026
Pricing accuracy: Official sources only (directly fetched)
Status: Ready for review (not yet committed)
🔗 Referenced by
- 📅Journal Entry - June 17, 20262026-06-17T00:00:00.000Z
- 📚HOW-TO: Build a Multi-Model Routing Layer for AI Applications2026-06-17T00:00:00.000Z
- 📚Wiki Index2026-06-17T00:00:00.000Z
- 📅Journal Entry - June 10, 20262026-06-10T00:00:00.000Z
- 🔬Claude Opus 4.8: Agentic Coding, Honesty, and Dynamic Workflows2026-05-28T00:00:00.000Z
- 📅Journal Entry - May 26, 20262026-05-26T00:00:00.000Z
- 🔬2026 Gartner Magic Quadrant for Enterprise AI Coding Agents: Market Map, Vendor Analysis, and Strategic Implications2026-05-26T00:00:00.000Z
- 🔬NVIDIA GPU Evolution: 2007-2026 Datacenter Architectures & Performance Scaling2026-05-11T00:00:00.000Z
- 🔬NVIDIA vs AMD GPUs: ROCm Ecosystem Maturity & Datacenter Competitive Landscape (2026)2026-05-11T00:00:00.000Z
- 📅Journal Entry - May 8, 20262026-05-08T00:00:00.000Z
- 📅Journal Entry - May 5, 20262026-05-05T00:00:00.000Z
- 📅Journal Entry - May 4, 20262026-05-04T00:00:00.000Z
- 📅Journal Entry - May 1, 20262026-05-01T00:00:00.000Z
- 📅Journal Entry - April 29, 20262026-04-29T00:00:00.000Z
- 📚Agentic Coding