feat: update model pricing with Claude 4.5 and version-specific rates#182
feat: update model pricing with Claude 4.5 and version-specific rates#182BenoitdeKersabiec wants to merge 1 commit intoMaciek-roboblog:mainfrom
Conversation
Update pricing to reflect current Anthropic pricing as of January 2025: - Add Claude 4.5 models (Opus, Sonnet, Haiku) with new lower rates - Add Claude 4 models with distinct pricing from 4.5 - Add Claude 3.5 models (Haiku, Sonnet) - Refactor fallback logic to detect model versions (4.5, 4, 3.5, 3) - Extract helper methods for cleaner code Pricing source: https://platform.claude.com/docs/en/about-claude/pricing
📝 WalkthroughWalkthroughThe pricing module is expanded with fallback pricing data for new Claude model variants (opus-4-5, sonnet-4-5, haiku-4-5, opus-4, sonnet-4, and claude-3 variants). Two internal helper methods are added to manage cache pricing defaults and fallback pricing lookup, with refactored pricing retrieval logic to use these helpers. Changes
Estimated code review effort🎯 3 (Moderate) | ⏱️ ~25 minutes Poem
🚥 Pre-merge checks | ✅ 2 | ❌ 1❌ Failed checks (1 warning)
✅ Passed checks (2 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing touches
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 2
🤖 Fix all issues with AI agents
In `@src/claude_monitor/core/pricing.py`:
- Around line 237-248: The is_4 detection can miss model names that end with
"-4" (e.g., "claude-opus-4"); update the is_4 boolean in pricing.py to also
check for model_lower.endswith("-4") (keeping the existing is_4_5 precedence) so
that the block that returns self.FALLBACK_PRICING["opus-4"] correctly handles
trailing "-4" names; change the is_4 assignment (the variable named is_4 that
uses model_lower, and the subsequent if "opus" branch that returns opus-4) to
include this endswith check.
- Around line 29-108: FALLBACK_PRICING contains an outdated rate for the
"haiku-3-5" key; update the "haiku-3-5" entry in the FALLBACK_PRICING dict to
use input: 1.0 and output: 5.0 (and adjust related cache_creation/cache_read if
you want them to match the 4.5 haiku multipliers) so it matches the current
Claude 3.5 Haiku pricing; locate the FALLBACK_PRICING constant and change the
"haiku-3-5" values accordingly (or add a comment explaining why the older 0.8/4
values are intentionally retained if you choose not to update).
🧹 Nitpick comments (1)
src/claude_monitor/core/pricing.py (1)
197-202: Avoid mutating the input dictionary.This method modifies the
pricingdict in place. Sinceself.pricingvalues are direct references toFALLBACK_PRICINGentries, any mutation would affect the class-level constant. Currently safe because all entries have cache keys, but fragile if custom pricing is provided without them.♻️ Return a new dict instead of mutating
def _ensure_cache_pricing(self, pricing: Dict[str, float]) -> Dict[str, float]: + result = pricing.copy() - if "cache_creation" not in pricing: - pricing["cache_creation"] = pricing["input"] * 1.25 - if "cache_read" not in pricing: - pricing["cache_read"] = pricing["input"] * 0.1 - return pricing + if "cache_creation" not in result: + result["cache_creation"] = result["input"] * 1.25 + if "cache_read" not in result: + result["cache_read"] = result["input"] * 0.1 + return result
| FALLBACK_PRICING: Dict[str, Dict[str, float]] = { | ||
| "opus": { | ||
| "opus-4-5": { | ||
| "input": 5.0, | ||
| "output": 25.0, | ||
| "cache_creation": 6.25, | ||
| "cache_read": 0.5, | ||
| }, | ||
| "sonnet-4-5": { | ||
| "input": 3.0, | ||
| "output": 15.0, | ||
| "cache_creation": 3.75, | ||
| "cache_read": 0.3, | ||
| }, | ||
| "haiku-4-5": { | ||
| "input": 1.0, | ||
| "output": 5.0, | ||
| "cache_creation": 1.25, | ||
| "cache_read": 0.1, | ||
| }, | ||
| "opus-4": { | ||
| "input": 15.0, | ||
| "output": 75.0, | ||
| "cache_creation": 18.75, | ||
| "cache_read": 1.5, | ||
| }, | ||
| "sonnet": { | ||
| "sonnet-4": { | ||
| "input": 3.0, | ||
| "output": 15.0, | ||
| "cache_creation": 3.75, | ||
| "cache_read": 0.3, | ||
| }, | ||
| "haiku": { | ||
| "haiku-3-5": { | ||
| "input": 0.8, | ||
| "output": 4.0, | ||
| "cache_creation": 1.0, | ||
| "cache_read": 0.08, | ||
| }, | ||
| "sonnet-3-5": { | ||
| "input": 3.0, | ||
| "output": 15.0, | ||
| "cache_creation": 3.75, | ||
| "cache_read": 0.3, | ||
| }, | ||
| "opus-3": { | ||
| "input": 15.0, | ||
| "output": 75.0, | ||
| "cache_creation": 18.75, | ||
| "cache_read": 1.5, | ||
| }, | ||
| "sonnet-3": { | ||
| "input": 3.0, | ||
| "output": 15.0, | ||
| "cache_creation": 3.75, | ||
| "cache_read": 0.3, | ||
| }, | ||
| "haiku-3": { | ||
| "input": 0.25, | ||
| "output": 1.25, | ||
| "cache_creation": 0.3, | ||
| "cache_read": 0.03, | ||
| }, | ||
| "opus": { | ||
| "input": 5.0, | ||
| "output": 25.0, | ||
| "cache_creation": 6.25, | ||
| "cache_read": 0.5, | ||
| }, | ||
| "sonnet": { | ||
| "input": 3.0, | ||
| "output": 15.0, | ||
| "cache_creation": 3.75, | ||
| "cache_read": 0.3, | ||
| }, | ||
| "haiku": { | ||
| "input": 1.0, | ||
| "output": 5.0, | ||
| "cache_creation": 1.25, | ||
| "cache_read": 0.1, | ||
| }, | ||
| } |
There was a problem hiding this comment.
🧩 Analysis chain
🌐 Web query:
Claude API pricing January 2025 official documentation
💡 Result:
As of January 2025, Anthropic’s official Claude API documentation lists token-based pricing (per 1M tokens) and related multipliers. Key published rates:
- Claude Sonnet (mid‑tier): Input $3 / 1M, Output $15 / 1M. [1][2]
- Claude Haiku (cost‑efficient): Input $0.80 / 1M, Output $4 / 1M. [1][2]
- Claude Opus (highest capability): Input $15 / 1M, Output $75 / 1M. [2]
Other documented details:
- Prompt‑caching and cache read/write multipliers apply (cache writes/read charged differently). [2]
- Batch (asynchronous) API: typically a 50% discount on input/output token rates for supported models. [1]
- Long‑context (1M token window) premium rates apply for >200K input tokens on Sonnet 4/4.5 (higher per‑token rates). [1]
Sources: Anthropic / Claude official docs. [1] https://docs.claude.com/en/docs/about-claude/pricing [2] https://docs.anthropic.com/en/docs/about-claude/models/all-models
🌐 Web query:
Claude 4.5 opus sonnet haiku pricing per million tokens 2025
💡 Result:
Here are Anthropic’s Claude 4.5 (Opus / Sonnet / Haiku) API prices per million tokens (USD):
- Opus 4.5 — Input: $5 / 1M tokens; Output: $25 / 1M. [1][2]
- Sonnet 4.5 — Input: $3 / 1M (prompts >200K: $6 / 1M); Output: $15 / 1M (prompts >200K: $22.50 / 1M). [2][3]
- Haiku 4.5 — Input: $1 / 1M; Output: $5 / 1M. [2][3]
Sources: Anthropic product/news and pricing pages. [1] Anthropic announcement; [2] Claude pricing page; [3] Claude docs (models & pricing).
🌐 Web query:
Claude 4 opus sonnet pricing input output per million tokens
💡 Result:
Summary (per‑million‑token API pricing):
- Claude Opus 4 / Opus 4.1 — Input: $15 / MTok; Output: $75 / MTok. [1][2]
- Claude Opus 4.5 — Input: $5 / MTok; Output: $25 / MTok. [2][3]
- Claude Sonnet 4 (and Sonnet 3.7) — Input: $3 / MTok; Output: $15 / MTok. [1][2]
- Claude Sonnet 4.5 — Typical: Input $3 / MTok and Output $15 / MTok for prompts ≤200K tokens; higher rates apply for prompts >200K (Input $6 / MTok, Output $22.50 / MTok). [2][3]
Sources: Anthropic/Claude pricing pages and docs. [1] [2] [3]
References:
[1] Anthropic models overview / pricing. (docs.anthropic.com)
[2] Claude pricing page (official). (claude.com)
[3] Recent pricing summary / blog posts (Opus/Sonnet 4.5 announcements). (website.claude.com)
Citations:
- 1: https://docs.anthropic.com/en/docs/about-claude/models/all-models?utm_source=openai
- 2: https://claude.com/redirect/claudeai.v1.dd735e58-de72-49a7-adb7-3bc468bf8560/pricing?utm_source=openai
- 3: https://website.claude.com/pricing?utm_source=openai
🌐 Web query:
Claude 3.5 haiku sonnet pricing input output per million tokens
💡 Result:
Short answer — per Anthropic’s docs (current model pages):
- Claude 3.5 Sonnet: $3 /M input tokens; $15 /M output tokens. [1][2]
- Claude 3.5 Haiku: $0.80 /M input tokens; $4 /M output tokens. [1][2]
Note: some news reporting (Nov 2024) states Claude 3.5 Haiku was later listed at $1 /M input and $5 /M output; you may see either pair in different pages. [3]
Sources: Anthropic model/pricing docs and Claude pricing page. [1][2][3]
🌐 Web query:
Claude 3 opus sonnet haiku pricing input output per million tokens
💡 Result:
Summary (price per 1M tokens, USD):
- Claude Opus (Claude 3 / Opus series): Input $15 / MTok — Output $75 / MTok. [1][2]
- Claude Sonnet (Claude 3 Sonnet family): Input $3 / MTok — Output $15 / MTok. [1][2]
- Claude Haiku (Claude 3 family): depends on sub‑version — e.g. Haiku 4.5: Input $1 / MTok — Output $5 / MTok; older Haiku variants: Haiku 3.5 ≈ $0.80 input / $4 output, Haiku 3 ≈ $0.25 input / $1.25 output. [1][2]
Notes: Anthropic’s docs also list prompt‑caching multipliers (5‑min and 1‑hr cache write rates and cache read rates) that affect billed rates; see the cited pricing table for details. [1][2]
Sources: Anthropic / Claude docs pricing tables. [1] [2]
🌐 Web query:
Claude prompt caching pricing cache write cache read multiplier per million tokens
💡 Result:
Short answer:
-
Multiplier: cache write = 1.25× base input price (25% higher); cache read = 0.10× base input price (10% of base). [1][2]
-
Example prices (per MTok = per million tokens) reported by Anthropic for Claude models:
- Claude 3.5 Sonnet — Input $3 / MTok; Cache write $3.75 / MTok; Cache read $0.30 / MTok. [1]
- Claude 3 Opus — Input $15 / MTok; Cache write $18.75 / MTok; Cache read $1.50 / MTok. [1]
- Claude 3 Haiku — Input $0.25 / MTok; Cache write $0.30 / MTok; Cache read $0.03 / MTok. [1]
- (Docs also show 5‑min vs 1‑hour write multipliers where 1‑hour writes = 2× base input price). [2]
Sources: Anthropic prompt-caching blog and Claude pricing docs. [1][2]
Pricing structure verified against official documentation and correctly implemented.
All base input/output prices match Anthropic's official Claude API pricing as of January 2025:
- 4.5 models: Opus ($5/$25), Sonnet ($3/$15), Haiku ($1/$5) ✓
- 4 models: Opus ($15/$75), Sonnet ($3/$15) ✓
- 3.5 models: Sonnet ($3/$15), Haiku ($0.80/$4) ✓
- 3 models: Opus ($15/$75), Sonnet ($3/$15), Haiku ($0.25/$1.25) ✓
Cache pricing multipliers are correctly applied (1.25× for cache writes, 0.10× for cache reads).
Minor note: The haiku-3-5 entry uses $0.80/$4 per official documentation, but Anthropic updated Claude 3.5 Haiku pricing to $1/$5 in November 2024. Consider updating haiku-3-5 to match the current rate, or document why the earlier pricing is retained. The generic haiku fallback correctly defaults to the current 4.5 pricing ($1/$5).
🤖 Prompt for AI Agents
In `@src/claude_monitor/core/pricing.py` around lines 29 - 108, FALLBACK_PRICING
contains an outdated rate for the "haiku-3-5" key; update the "haiku-3-5" entry
in the FALLBACK_PRICING dict to use input: 1.0 and output: 5.0 (and adjust
related cache_creation/cache_read if you want them to match the 4.5 haiku
multipliers) so it matches the current Claude 3.5 Haiku pricing; locate the
FALLBACK_PRICING constant and change the "haiku-3-5" values accordingly (or add
a comment explaining why the older 0.8/4 values are intentionally retained if
you choose not to update).
| is_4_5 = "4-5" in model_lower or "4.5" in model_lower | ||
| is_4 = not is_4_5 and ( | ||
| "4-" in model_lower or "-4-" in model_lower or "4.1" in model_lower | ||
| ) | ||
| is_3_5 = "3-5" in model_lower or "3.5" in model_lower | ||
|
|
||
| if "opus" in model_lower: | ||
| return self.FALLBACK_PRICING["opus"] | ||
| if is_4_5: | ||
| return self.FALLBACK_PRICING["opus-4-5"] | ||
| if is_4: | ||
| return self.FALLBACK_PRICING["opus-4"] | ||
| return self.FALLBACK_PRICING["opus-3"] |
There was a problem hiding this comment.
Version detection may miss models ending in "-4".
The is_4 check looks for "4-" or "-4-" in the model name, but won't match model names that end with -4 (e.g., "claude-opus-4", "some-opus-4"). While currently mapped models are handled in self.pricing, future model variations or API name differences could fall through incorrectly.
🔧 Suggested fix to handle trailing -4
- is_4 = not is_4_5 and (
- "4-" in model_lower or "-4-" in model_lower or "4.1" in model_lower
- )
+ is_4 = not is_4_5 and (
+ "4-" in model_lower or "-4-" in model_lower or
+ model_lower.endswith("-4") or "4.1" in model_lower
+ )📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
| is_4_5 = "4-5" in model_lower or "4.5" in model_lower | |
| is_4 = not is_4_5 and ( | |
| "4-" in model_lower or "-4-" in model_lower or "4.1" in model_lower | |
| ) | |
| is_3_5 = "3-5" in model_lower or "3.5" in model_lower | |
| if "opus" in model_lower: | |
| return self.FALLBACK_PRICING["opus"] | |
| if is_4_5: | |
| return self.FALLBACK_PRICING["opus-4-5"] | |
| if is_4: | |
| return self.FALLBACK_PRICING["opus-4"] | |
| return self.FALLBACK_PRICING["opus-3"] | |
| is_4_5 = "4-5" in model_lower or "4.5" in model_lower | |
| is_4 = not is_4_5 and ( | |
| "4-" in model_lower or "-4-" in model_lower or | |
| model_lower.endswith("-4") or "4.1" in model_lower | |
| ) | |
| is_3_5 = "3-5" in model_lower or "3.5" in model_lower | |
| if "opus" in model_lower: | |
| if is_4_5: | |
| return self.FALLBACK_PRICING["opus-4-5"] | |
| if is_4: | |
| return self.FALLBACK_PRICING["opus-4"] | |
| return self.FALLBACK_PRICING["opus-3"] |
🤖 Prompt for AI Agents
In `@src/claude_monitor/core/pricing.py` around lines 237 - 248, The is_4
detection can miss model names that end with "-4" (e.g., "claude-opus-4");
update the is_4 boolean in pricing.py to also check for
model_lower.endswith("-4") (keeping the existing is_4_5 precedence) so that the
block that returns self.FALLBACK_PRICING["opus-4"] correctly handles trailing
"-4" names; change the is_4 assignment (the variable named is_4 that uses
model_lower, and the subsequent if "opus" branch that returns opus-4) to include
this endswith check.
2 participants
Update pricing to reflect current Anthropic pricing as of January 2025:
Pricing source: https://platform.claude.com/docs/en/about-claude/pricing
Summary by CodeRabbit
Release Notes
New Features
Improvements
✏️ Tip: You can customize this high-level summary in your review settings.