Alternatives to GPT-5.3-Codex

These models are alternatives by measurable coding evidence: they share the current closed-model classificationand are closest to GPT-5.3-Codex's SWE-bench score. Similarity here does not mean identical capabilities, licence terms, or output quality.

The target model

SWE-bench
78.0%
Cheapest input
$1.75
Cheapest output
$14.00
Hosts
4
AlternativeSWE-benchInput $/MTokOutput $/MTokHostsContextTradeoff
gpt-5.4 (<272K context length)78.2%$2.50$15.006higher input price
GLM 5.278.7%$0.50$2.0066203klower input price
Claude Opus 4.678.7%$5.00$25.00101.0Mhigher input price
Gemini 3.5 Flash78.8%$0.75$4.504lower input price
Gemini 3.1 Pro Preview78.8%$1.00$6.0041.0Mlower input price
Gemini 3.6 Flash79.6%$0.38$1.884lower input price
Claude Sonnet 579.6%$2.00$10.0010higher input price
Claude Opus 4.576.4%$5.00$25.0010200khigher input price
gpt-5.275.8%$1.75$14.004closest measured match
Gemini 3.1 Pro Preview Custom Tools75.6%$2.00$12.0021.0Mhigher input price
Gemini 3 Flash Preview75.4%$0.25$1.504lower input price
Claude Sonnet 4.675.2%$3.00$15.00101.0Mhigher 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.