June 2026 AI Model Releases: GPT-5.4 Family + Claude Fable 5
Six new AI models launched in June 2026. OpenAI released the GPT-5.4 family — four models spanning $0.20 to $30 per million tokens. Anthropic shipped Claude Fable 5, its most capable widely-released model, and the restricted Claude Mythos 5 for security research. Here's everything you need to know about pricing, capabilities, and which model fits your workload.
Last updated: June 27, 2026 · All prices per 1M tokens
Key Takeaways
- GPT-5.4 nano is OpenAI's cheapest model ever at $0.20/$1.25 — 84% cheaper than GPT-5.4
- GPT-5.4 at $2.50/$15 replaces GPT-4o as the mid-tier workhorse with 400K context
- Claude Fable 5 at $10/$50 is Anthropic's new flagship — 1M context, adaptive thinking
- GPT-5.4 Pro at $30/$180 matches GPT-5.5 Pro pricing for maximum capability
- The budget tier now ranges from $0.08 (GPT-oss 20B) to $0.75 (GPT-5.4 mini)
🟢 OpenAI GPT-5.4 Family
OpenAI released four GPT-5.4 models in June 2026, filling the gap between GPT-5 mini and GPT-5.5. The family spans every price point — from the ultra-cheap nano to the premium Pro variant.
GPT-5.4
The direct successor to GPT-4o. Same $2.50 input price but 50% more output tokens and 3× the context window (400K vs 128K). Best for: chatbots, content generation, code assistance, and general reasoning tasks.
GPT-5.4 mini
70% cheaper than GPT-5.4 with comparable quality for straightforward tasks. Ideal for: classification, summarization, data extraction, and high-volume API calls where cost matters more than peak intelligence.
GPT-5.4 nano
At $0.20 per million input tokens, GPT-5.4 nano is 92% cheaper than GPT-5.4. Still offers 400K context — unusual for a budget model. Best for: autocomplete, simple Q&A, routing, and embedding-like tasks at scale.
GPT-5.4 Pro
The top-tier GPT-5.4 model. Same price as GPT-5.5 Pro but with 400K context (vs 1.05M). For complex multi-step reasoning, agentic coding, and tasks where output quality is paramount and budget is flexible.
🟣 Anthropic Claude Fable 5
Claude Fable 5 launched June 9, 2026 as Anthropic's most capable widely-released model. It slots above Opus 4.8 with a new tokenizer that produces ~30% more tokens per text compared to earlier Claude models.
Claude Fable 5
Available on Claude API, Amazon Bedrock, Google Cloud, and Microsoft Foundry. Uses the new tokenizer from Opus 4.7 (~30% more tokens per text). Notable: no extended thinking support — uses adaptive thinking instead. Best for: complex reasoning, long-horizon agentic work, and tasks requiring maximum intelligence.
Claude Mythos 5 ⚠️ Restricted
Anthropic's restricted model for cybersecurity and biology research. Same pricing as Fable 5 but access is limited to approved customers through Project Glasswing. Requires accepting a 30-day data retention policy for safety monitoring. The same model is offered as Claude Fable 5 with additional safeguards for public use.
Complete Pricing Comparison
All 6 new models alongside existing competitors. Prices per 1M tokens. Sorted by input price.
| Model | Provider | Tier | Input | Output | Context | Status |
|---|---|---|---|---|---|---|
| GPT-5.4 nano | OpenAI | Budget | $0.20 | $1.25 | 400K | NEW |
| GPT-5.4 mini | OpenAI | Budget | $0.75 | $4.50 | 400K | NEW |
| GPT-5 | OpenAI | Mid | $1.25 | $10.00 | 272K | Active |
| GPT-5.4 | OpenAI | Mid | $2.50 | $15.00 | 400K | NEW |
| GPT-4o | OpenAI | Mid | $2.50 | $10.00 | 128K | Active |
| Claude Sonnet 4.6 | Anthropic | Mid | $3.00 | $15.00 | 1M | Active |
| GPT-5.5 | OpenAI | Premium | $5.00 | $30.00 | 1.05M | Active |
| Claude Opus 4.8 | Anthropic | Premium | $5.00 | $25.00 | 1M | Active |
| Gemini 3.1 Pro | Mid | $2.00 | $12.00 | 1M | Active | |
| Gemini 3.5 Flash | Mid | $1.50 | $9.00 | 1M | Active | |
| Claude Fable 5 | Anthropic | Premium | $10.00 | $50.00 | 1M | NEW |
| Claude Mythos 5 | Anthropic | Premium | $10.00 | $50.00 | 1M | NEW ⚠️ |
| DeepSeek V4 Pro | DeepSeek | Budget | $0.44 | $0.87 | 1M | Active |
| DeepSeek V4 Flash | DeepSeek | Budget | $0.14 | $0.28 | 1M | Active |
Cost Per Task: What You'll Actually Pay
Real-world cost estimates for common workloads. Based on typical token counts per task.
| Task | GPT-5.4 nano | GPT-5.4 mini | GPT-5.4 | Claude Fable 5 | GPT-5.5 |
|---|---|---|---|---|---|
| Simple chat (1K→500 tokens) | $0.00083 | $0.00300 | $0.01000 | $0.03500 | $0.02000 |
| Long doc summary (10K→2K tokens) | $0.00450 | $0.01650 | $0.05500 | $0.20000 | $0.11000 |
| Code generation (5K→3K tokens) | $0.00475 | $0.01725 | $0.05750 | $0.20000 | $0.11500 |
| Agent step (20K→5K tokens) | $0.01025 | $0.03750 | $0.12500 | $0.45000 | $0.25000 |
| RAG query (15K→3K tokens) | $0.00675 | $0.02475 | $0.08250 | $0.30000 | $0.16500 |
💡 Key insight: GPT-5.4 nano costs less than $0.01 for most single tasks. At scale (1M requests/month), the difference between nano and Fable 5 is $34,170/month.
Which Model Should You Use?
💰 Best Budget
GPT-5.4 nano at $0.20/$1.25
84% cheaper than GPT-5.4. 400K context. Best for high-volume, simple tasks.
⚖️ Best Value
GPT-5.4 mini at $0.75/$4.50
70% cheaper than GPT-5.4 with near-identical quality for most tasks.
🧠 Best Intelligence
Claude Fable 5 at $10/$50
Anthropic's most capable model. 1M context. Adaptive thinking. Complex reasoning.
⚡ Best Mid-Tier
GPT-5.4 at $2.50/$15
The new GPT-4o. Same price, 3× context, better quality. General workhorse.
Migration Guide: Upgrading to New Models
| If You're Using | Upgrade To | Cost Change | Context Change |
|---|---|---|---|
| GPT-4o ($2.50/$10) | GPT-5.4 ($2.50/$15) | +50% output | 128K → 400K (3×) |
| GPT-4o mini ($0.15/$0.60) | GPT-5.4 mini ($0.75/$4.50) | 5× more expensive | 128K → 400K (3×) |
| GPT-5 ($1.25/$10) | GPT-5.4 ($2.50/$15) | 2× more expensive | 272K → 400K (1.5×) |
| Claude Opus 4.8 ($5/$25) | Claude Fable 5 ($10/$50) | 2× more expensive | 1M → 1M (same) |
| Claude Sonnet 4.6 ($3/$15) | GPT-5.4 ($2.50/$15) | 17% cheaper input | 1M → 400K (smaller) |
⚠️ Note: Claude Fable 5 uses a new tokenizer that produces ~30% more tokens per text than earlier Claude models. Your token counts will be higher, which affects cost calculations even though the per-token price is the same as Mythos 5.
Calculate Your Exact Costs
Enter your token usage and see exactly what each model costs for your workload.
Open Cost Calculator →Related Tools
Get APIpulse Pro — $29
Save scenarios, PDF reports, optimization tips, and cost alerts for 48 AI models. One-time payment, no subscription.
Unlock Pro →Frequently Asked Questions
How much does GPT-5.4 cost?
GPT-5.4 costs $2.50 per 1M input tokens and $15.00 per 1M output tokens. It has a 400K token context window. The cheaper GPT-5.4 mini costs $0.75/$4.50 and the ultra-budget GPT-5.4 nano costs $0.20/$1.25.
What is Claude Fable 5 and how much does it cost?
Claude Fable 5 is Anthropic's most capable widely released model, launched June 9, 2026. It costs $10 per 1M input tokens and $50 per 1M output tokens with a 1M token context window and 128K max output. It uses adaptive thinking (always on) and the new tokenizer from Opus 4.7.
What is Claude Mythos 5?
Claude Mythos 5 is Anthropic's restricted-access model available through Project Glasswing. It costs $10/$50 per MTok (same as Fable 5) with a 1M context window. It's designed for cybersecurity and biology research. Access is invitation-only with a 30-day data retention policy. The same model is available as Claude Fable 5 with additional safeguards.
Which new model is cheapest?
GPT-5.4 nano is the cheapest new model at $0.20 per 1M input tokens and $1.25 per 1M output tokens. That's 84% cheaper than GPT-5.4 on input and 72% cheaper on output. For comparison, DeepSeek V4 Flash ($0.14/$0.28) is still cheaper on input but GPT-5.4 nano has a much larger 400K context window.
Is GPT-5.4 Pro worth the premium over GPT-5.5?
GPT-5.4 Pro costs $30/$180 per MTok — the same as GPT-5.5 Pro. Both have 400K+ context windows. If you need the absolute highest OpenAI capability, GPT-5.5 Pro ($30/$180, 1.05M context) offers more context. For most use cases, GPT-5.4 ($2.50/$15) provides excellent value at 12× less cost.
All pricing data sourced from official provider documentation as of June 27, 2026. Prices may change. Use the APIpulse Cost Calculator for the latest estimates.