Alternatives to Claude Haiku 4.5
These models are alternatives by measurable coding evidence: they share the current closed-model classificationand are closest to Claude Haiku 4.5's SWE-bench score. Similarity here does not mean identical capabilities, licence terms, or output quality.
The target model
SWE-bench
66.6%
Cheapest input
$1.00
Cheapest output
$5.00
Hosts
8
| Alternative | SWE-bench | Input $/MTok | Output $/MTok | Hosts | Context | Tradeoff |
|---|---|---|---|---|---|---|
| GPT-4o-mini (2024-07-18) | 64.8% | $0.15 | $0.60 | 2 | 128k | lower input price |
| gpt-5 | 69.0% | $1.25 | $10.00 | 4 | — | higher input price |
| GLM 4.7 | 69.4% | $0.38 | $1.74 | 24 | 203k | lower input price |
| gpt-5.4-nano | 69.8% | $0.20 | $1.25 | 4 | — | lower input price |
| gpt-5.1 | 69.8% | $1.25 | $10.00 | 4 | — | higher input price |
| Claude Sonnet 4.5 | 70.0% | $3.00 | $15.00 | 10 | 1.0M | higher input price |
| Gemini 3.1 Flash Lite Preview | 62.8% | $0.13 | $0.75 | 3 | 1.0M | lower input price |
| Grok 4.3 | 71.4% | $1.25 | $2.50 | 2 | 1.0M | higher input price |
| GPT-4o (2024-08-06) | 71.9% | $2.50 | $10.00 | 4 | 128k | higher coding score, higher input price |
| GLM 5 | 72.1% | $0.60 | $1.92 | 31 | 203k | higher score, no higher input price |
| gpt-5-mini | 60.8% | $0.25 | $2.00 | 4 | — | cheaper, lower coding score |
| GPT-5.2-Codex | 72.4% | $1.75 | $14.00 | 3 | 400k | higher coding score, higher input price |
How alternatives are selected
Alternatives must have a benchmark score and the same current open/closed classification as the target. They are sorted by score distance, then by price/performance. The table compares hosted API economics; it does not establish licence compatibility or self-hosting rights.
Check the model licence, provider Terms, context limits, tool support, and output quality before switching a production workload.