🔥 Limited time: Pro lifetime access $19 — price goes up July 12 →
Claude Sonnet 5 launches at $2/$10. GPT-4.1 replaces GPT-4o at 20% cheaper. Mistral Small 4 enters the ultra-budget tier. Here's where every AI API stands — 58 models, 10 providers, real quality data.
July brought the biggest wave of model launches and deprecations since the Claude 4 shutdown. Claude Sonnet 5 arrived at aggressive intro pricing, OpenAI replaced the entire GPT-4o family with GPT-4.1, and Mistral entered the ultra-budget tier.
Anthropic launched Claude Sonnet 5 at $2/$10 per 1M tokens (intro pricing through Aug 31, regular $3/$15). With a 1M context window and quality matching Claude Sonnet 4.6, it's the best mid-tier value right now. The intro price makes it cheaper than GPT-4.1 ($2/$8) on output tokens — a first for Anthropic.
OpenAI officially deprecated GPT-4o and GPT-4o mini. The GPT-4.1 family is 20% cheaper with 8× larger context (1M vs 128K). GPT-4.1 ($2/$8), GPT-4.1 mini ($0.40/$1.60), and GPT-4.1 nano ($0.10/$0.40) span the full price range. Note: GPT-4.1 nano is already marked for shutdown on Oct 23, 2026.
OpenAI deprecated GPT-5 ($1.25/$10) and GPT-5 mini ($0.25/$2) with a shutdown date of Dec 11, 2026. Migrate to GPT-5.5 ($5/$30) for premium or GPT-5.4 mini ($0.40/$1.60) for budget. GPT-oss 20B at $0.08/$0.35 remains the cheapest OpenAI-hosted option.
Mistral launched Small 4 at $0.15/$0.60 per 1M tokens with 128K context. It's price-competitive with DeepSeek V4 Flash ($0.14/$0.28) and significantly cheaper than Jamba Mini ($0.20/$0.40). A strong option for EU data residency requirements.
Together.ai delisted Llama 4 Scout and Maverick from serverless inference in July 2026. Llama 3.1 70B and 8B were also removed. The recommended replacement is Llama 3.3 70B at $1.04/$1.04. These models may still be available via dedicated endpoints.
Every active AI API model, sorted by output price. Deprecated models marked in red. Prices per million tokens.
| Model | Provider | Tier | Input $/1M | Output $/1M | Context | vs Claude 4 Opus |
|---|
Prices verified Jul 8, 2026. Claude 4 Opus was $15.00 input / $75.00 output per 1M tokens.
10 providers, each with different strengths. Here's who leads in price, quality, and reliability.
Cheapest: Haiku 4.5 — $1.00/$5.00
4 active models · Opus 4.8, Sonnet 5, Sonnet 4.6, Haiku 4.5
Highest quality (97% of Claude 4 Opus). Sonnet 5 at intro pricing ($2/$10) is the best mid-tier value. Best for complex reasoning, coding, and creative work.
Cheapest: V4 Flash — $0.14/$0.28
2 active models · V4 Pro, V4 Flash
Price leader. V3.2 deprecated Jul 7. 99% cheaper than Claude 4 Opus. Excellent for batch processing, prototyping, and cost-sensitive workloads.
Cheapest: GPT-oss 20B — $0.08/$0.35
8 active models · GPT-5.5, 5.4-mini, 4.1, 4.1-mini, oss
Broadest model range. GPT-4.1 replaces GPT-4o at 20% cheaper. GPT-5 and GPT-5 mini deprecated (shutdown Dec 11). GPT-oss models are cheapest.
Cheapest: Gemini 2.5 Flash-Lite — $0.10/$0.40
7 active models · 3.5 Flash, 3.1 Pro, 2.5 Pro, 2.5 Flash
Best context window (1M tokens). Gemini 2.0 Flash/Flash-Lite deprecated. Great for long-document processing and multimodal tasks.
Cheapest: Small 4 — $0.15/$0.60
3 active models · Large 3, Medium 3.5, Small 4
Small 4 is new ultra-budget contender. European provider with strong privacy focus. Price-competitive with DeepSeek for EU data residency.
Cheapest: Llama 3.3 70B — $1.04/$1.04
1 active model · Llama 3.3 70B
Llama 4 Scout/Maverick delisted from serverless Jul 2026. Llama 3.1 70B/8B also removed. Llama 3.3 70B is the recommended open-source option.
Cheapest: Command R — $0.50/$1.50
3 active models · Command A, R+, R
Enterprise-focused with strong RAG capabilities. Command A is competitive with GPT-5 for retrieval-augmented generation tasks.
Cheapest: Grok Build 0.1 — $1.00/$2.00
2 active models · Grok 4.3, Grok Build 0.1
Grok Build 0.1 repriced to $1.00/$2.00. Grok 4.3 at $1.25/$2.50 with 1M context. Good for developers in the X/Twitter ecosystem.
Cheapest: Jamba Mini — $0.20/$0.40
2 active models · Jamba 1.7 Large, Jamba Mini
Jamba 1.5 Large deprecated (superseded by 1.7). Hybrid SSM+Transformer architecture for efficient 256K context processing.
Cheapest: Kimi K2.6 — $0.95/$4.00
1 active model · Kimi K2.6
Chinese provider with 256K context. 10× more expensive than DeepSeek V4 Flash. Niche use case for Chinese language tasks.
There's no single "best" AI API. The right choice depends on your workload, budget, and quality requirements.
Complex reasoning, coding, creative work
Claude Opus 4.8
$5.00/$25.00 per 1M tokens
97% of Claude 4 Opus quality. The safest choice for quality-critical workloads. Worth the premium for important tasks.
General purpose, good quality at low cost
Claude Sonnet 5
$2.00/$10.00 per 1M tokens (intro)
Best mid-tier value right now. 1M context, quality matching Sonnet 4.6. Intro pricing through Aug 31 — regular $3/$15.
Batch processing, prototyping, high volume
DeepSeek V4 Flash
$0.14/$0.28 per 1M tokens
99% cheaper than Claude 4 Opus. Perfect for non-critical tasks, data processing, and rapid prototyping.
Mission-critical, enterprise SLA
GPT-4.1
$2.00/$8.00 per 1M tokens
Replaces GPT-4o at 20% cheaper with 8× larger context. OpenAI's enterprise infrastructure. Safe choice for production.
Document analysis, long conversations
Gemini 3.1 Pro
$2.00/$12.00 per 1M tokens
1M token context window. Best for processing long documents, codebases, and multi-turn conversations.
GDPR, data residency requirements
Mistral Small 4
$0.15/$0.60 per 1M tokens
European provider with strong privacy guarantees. New ultra-budget option. Best for EU-regulated industries.
If you were spending $500/month on Claude 4 Opus, here's what you'd pay with each alternative:
| Provider | Model | Monthly Cost | Savings | Quality |
|---|---|---|---|---|
| RETIRED | Claude 4 Opus | $500.00 | — | 100% |
| Anthropic | Claude Opus 4.8 | $165.00 | -$335 (67%) | 97% |
| Anthropic | Claude Sonnet 5 INTRO | $50.00 | -$450 (90%) | 90% |
| OpenAI | GPT-4.1 | $40.00 | -$460 (92%) | 88% |
| Gemini 2.5 Pro | $40.00 | -$460 (92%) | 85% | |
| DeepSeek | DeepSeek V4 Pro | $14.50 | -$486 (97%) | 82% |
| Mistral | Mistral Small 4 | $5.50 | -$495 (99%) | 70% |
| DeepSeek | DeepSeek V4 Flash | $4.70 | -$495 (99%) | 72% |
| Gemini 2.5 Flash-Lite | $1.75 | -$498 (99.6%) | 65% |
Based on ~33M tokens/month (1:3 input:output ratio). Quality scores from our internal benchmark suite. Sonnet 5 intro pricing through Aug 31.
Several models are approaching shutdown dates. Migrate before you lose access.
| Model | Provider | Shutdown Date | Replacement | Current Price |
|---|---|---|---|---|
| GPT-4.1 nano | OpenAI | Oct 23, 2026 | GPT-5.4 nano | $0.10/$0.40 |
| GPT-5 | OpenAI | Dec 11, 2026 | GPT-5.5 | $1.25/$10.00 |
| GPT-5 mini | OpenAI | Dec 11, 2026 | GPT-5.4 mini | $0.25/$2.00 |
| Claude Sonnet 4.6 | Anthropic | Deprecated | Claude Sonnet 5 | $3/$15 |
| DeepSeek V3.2 | DeepSeek | Deprecated | DeepSeek V4 Flash | N/A |
The AI API price war continues to accelerate. Here's what we expect:
After Aug 31, Claude Sonnet 5 returns to $3/$15. Still competitive with GPT-4.1 ($2/$8) on a total-cost basis due to 1M context. Expect Anthropic to hold this price through Q4 2026.
With GPT-4.1 nano shutting down Oct 23, OpenAI will launch GPT-5.4 nano as the replacement. Expect pricing at $0.10-$0.15 input — maintaining the ultra-budget tier.
The Dec 11 GPT-5/GPT-5 mini shutdown will trigger another mass migration. Developers should start testing alternatives now. GPT-5.4 mini ($0.40/$1.60) and DeepSeek V4 Pro ($0.44/$0.87) are the top destinations.
The budget tier is converging around $0.10-$0.20 input. DeepSeek V4 Flash ($0.14), Mistral Small 4 ($0.15), Jamba Mini ($0.20), and Gemini 2.5 Flash-Lite ($0.10) are all within 2× of each other. Quality differences will determine the winner, not price.
This report covers the landscape. Pro tells you exactly which model to use for YOUR workload — with code snippets, cost projections, and a migration plan.
Get Your Migration Plan — $19One-time payment. Lifetime access. No subscription. Flash sale ends Jul 12.