Comparison · AI Models & LLMs
GLM-5 vs Claude Haiku 4.5
The same method as our rankings: scores from the engine, sourced data, verified prices.
Details: Claude Haiku 4.5 →Details: GLM-5 →
Claude Haiku 4.5
Anthropic
Full sheet →- SWE-bench accuracy
- 66.6%
- Input cost per 1M tokens
- $1.00
- Output cost per 1M tokens
- $5.00
GLM-5
Z.ai
Full sheet →71.3/100
- SWE-bench accuracy
- 72.8%
- Context window
- 200k tokens
- Maximum output
- 128k tokens
- Input cost per 1M tokens
- $1.00
The numbers side by side
Each bar is the product's verified figure; on every row the longer bar is the higher of the two values.
- SWE-bench accuracyhigher is better
- Input cost per 1M tokenslower is better
- Cached input cost per 1M tokenslower is better
- Output cost per 1M tokenslower is better
Who wins where
- For codingGLM-5SWE-bench accuracy
- Cheaper to writeGLM-5Output cost per 1M tokens
GLM-5: open weights (downloadable)
Data sheet side by side
Performance
SWE-bench accuracy66.6%✓72.8%9.3% more
Memory and answer length
Context windowNot stated200k tokens
Maximum outputNot stated128k tokens
API prices
Input cost per 1M tokens$1.00$1.00
Cached input cost per 1M tokens✓$0.1050% less$0.20
Output cost per 1M tokens$5.00✓$3.2036% less
Size and openness
ParametersNot stated744 billion
Active parameters per tokenNot stated40 billion
Open weights (downloadable)Not statedYes
What a month costs
Change the volumes: the cost is worked out from the verified list prices above.
Claude Haiku 4.5$4.50per month · $54.00 per year
GLM-5$3.60per month · $43.20 per yearThe cheaper one — you save $0.90 per month
The scenarios are only starting points to edit, not data about your use. List prices without discounts, time-of-day rates or taxes.
Questions and answers
For coding: Claude Haiku 4.5 or GLM-5?
GLM-5: 72.8% against 66.6% (SWE-bench accuracy).
Cheaper to write: Claude Haiku 4.5 or GLM-5?
GLM-5: $3.20 against $5.00 (Output cost per 1M tokens).