This lesson's interactive features are locked, please to keep using them
Model Costs
Although you might be using DeepSeek V4.1 Flash Free in this course (or another model of your choice), it's not my daily driver for most coding work.
Click to play video
Every couple of weeks, it feels like a new SOTA (state-of-the-art) model drops. At the time of writing, some representative top-end models are:
| Developer | Model | Input Cost (1M Tokens) | Output Cost (1M Tokens) |
|---|---|---|---|
| OpenAI | gpt-6-astra |
$10.00 | $50.00 |
| OpenAI | gpt-5.6-sol |
$4.00 | $20.00 |
| Anthropic | claude-fable-5-1 |
$10.00 | $50.00 |
| Anthropic | claude-opus-5-5 |
$4.00 | $20.00 |
gemini-4-argon |
$2.00 | $10.00 |
Prices are in USD per 1 million tokens at standard API rates with uncached input and the base context tier, verified October 6, 2026 against OpenAI, Anthropic, Google, and Moonshot AI; other providers may charge different rates for the same model.
SOTA models, in addition to being powerful, tend to be expensive and slow(ish).
Input tokens are generally much less expensive than output tokens, and are cached aggressively, further reducing costs. That said, the volume of input tokens when coding tends to be massive, so they can still dominate the overall bill compared to output tokens.
Cheaper Models
Move down the price curve a bit and you usually also move up the speed curve. A few cheaper models worth knowing about:
| Developer | Model ID | Input Cost (1M Tokens) | Output Cost (1M Tokens) |
|---|---|---|---|
| OpenAI | gpt-5.6-luna |
$0.20 | $1.20 |
| Anthropic | claude-haiku-4-5 |
$1.00 | $5.00 |
gemini-3.8-flash |
$0.75 | $3.75 | |
| Moonshot AI | kimi-k2.6 |
$0.95 | $4.00 |
Argon's announced introductory pricing rises to $4/$20 after the introductory period, with no public API model ID confirmed yet; Gemini 3.8 Flash's introductory pricing lasts through December 31, 2026, then rises to $1.50/$7.50.