GPT-5.4 Mini vs Claude Haiku 4.5
Choose two or three models. Copy the URL to share your selection.
| Comparison | GPT-5.4 Minigpt-5.4-mini | Claude Haiku 4.5claude-haiku-4-5 |
|---|---|---|
| Type | Chat | Chat |
| Best for | Fast, low-cost replies and high-volume everyday tasks | Fast, low-cost replies and high-volume everyday tasks |
| TokenRex price | $0.75 in / $4.50 out per 1M tokens | $1.00 in / $5.00 out per 1M tokens |
| Cached input / 1M tokens | $0.075 per 1M tokens | $0.10 per 1M tokens |
| Cache write / 1M tokens | No extra fee | $1.25 per 1M tokens |
| Cached output | Not offered · standard output rate applies | Not offered · standard output rate applies |
| Market price | $0.75 in / $4.50 out per 1M tokens | $1.00 in / $5.00 out per 1M tokens |
| Context window (provider) | 400,000 tokens | 200,000 tokens |
| Input token limit (provider) | Within the context window | Within the context window |
| Maximum output | 128,000 tokens | 64,000 tokens |
| Knowledge cutoff | Aug 31, 2025 | Feb 2025 |
| Vendor input modalities | Not verified | Text, Image |
| Vendor output modalities | Not verified | Text |
| TokenRex endpoints | POST /v1/chat/completions | POST /v1/chat/completions · POST /v1/messages |
| TokenRex input | Text, Image/PDF content through the chat bridge | Text, Image/PDF content through the chat bridge |
| TokenRex output | Text | Text |
| Benchmarks | ||
| Benchmark evidence | 5 source-linked results | 1 source-linked results |
| SWE-Bench Pro (Public) | 54.4% Published reasoning effort: xhigh. SWE-Bench Pro (Public) | 57.7% | 54.4% | 52.4% | 45.7% | Published model-family result, not a test of TokenRex or its default effort. OpenAI - Introducing GPT-5.4 mini and nano | — |
| Terminal-Bench 2.0 | 60% Published reasoning effort: xhigh. Terminal-Bench 2.0 | 75.1% | 60.0% | 46.3% | 38.2% | Published model-family result, not a test of TokenRex or its default effort. OpenAI - Introducing GPT-5.4 mini and nano | — |
| Toolathlon | 42.9% Published reasoning effort: xhigh. Toolathlon | 54.6% | 42.9% | 35.5% | 26.9% | Published model-family result, not a test of TokenRex or its default effort. OpenAI - Introducing GPT-5.4 mini and nano | — |
| GPQA Diamond | 88% Published reasoning effort: xhigh. GPQA Diamond | 93.0% | 88.0% | 82.8% | 81.6% | Published model-family result, not a test of TokenRex or its default effort. OpenAI - Introducing GPT-5.4 mini and nano | — |
| OSWorld-Verified | 72.1% Published reasoning effort: xhigh. OSWorld-Verified | 75.0% | 72.1% | 39.0% | 42.0% | Published model-family result, not a test of TokenRex or its default effort. OpenAI - Introducing GPT-5.4 mini and nano | — |
| SWE-bench Verified | — | 73.3% Anthropic methodology footnote: simple bash+file-edit scaffold, averaged over 50 trials, no test-time compute, 128K thinking budget, default sampling, full 500-problem set, with a prompt addendum instructing heavy tool use. Haiku 4.5's catalog default is Extended thinking (no explicit low/medium/high effort knob), so this is treated as the base/default-thinking configuration, not a stripped-down no-thinking run. Anthropic · Introducing Claude Haiku 4.5 |
No overall score, winner or cost-versus-score frontier is calculated without comparable, attributable benchmark evidence. Different test configurations appear separately. Vendor capabilities are distinct from TokenRex API support.
— means no published value or no matching capability. Provider token limits are not tested gateway guarantees; TokenRex media controls and response formats are listed separately.
GPT-5.4 Mini
Sources and verification
Provider specifications describe the upstream model, not a guarantee of every feature through TokenRex. Prices come from the enabled catalog; benchmark results belong to the exact tested model.
- Provider model specifications
Vendor-reported
Last checked - OpenAI - Introducing GPT-5.4 mini and nano
Vendor-reported
Last checked
Research notes (1)
- Published provider limits; maximum-capacity requests have not been independently exercised through TokenRex. Thinking-level variants use the same provider model; reasoning and answer tokens share the output budget.
Claude Haiku 4.5
Sources and verification
Provider specifications describe the upstream model, not a guarantee of every feature through TokenRex. Prices come from the enabled catalog; benchmark results belong to the exact tested model.
- Provider model specifications
Vendor-reported
Last checked - Anthropic · Introducing Claude Haiku 4.5
Vendor-reported
Last checked
Research notes (1)
- Published provider limits; maximum-capacity requests have not been independently exercised through TokenRex. Thinking-level variants use the same provider model; reasoning and answer tokens share the output budget.
Popular comparisons
- GPT-6 Sol vs Claude Opus 5.5
- GPT-6 Sol vs Gemini 4 Argon
- Claude Opus 5.5 vs Gemini 4 Argon
- GPT-6.1 Sol vs GPT-6 Sol
- GPT-6 Astra vs Claude Sonnet 5
- GPT-6 Astra vs Gemini 3.1 Pro
- Claude Sonnet 5 vs Gemini 3.1 Pro
- Claude Opus 5.5 vs Claude Sonnet 5
- GPT-5.5 vs Claude Sonnet 5
- GPT-5.5 vs Gemini 2.5 Pro
- GPT-5 vs Gemini 2.5 Pro
- GPT-6 Luna vs Gemini 3.8 Flash
- Claude Haiku 4.5 vs Gemini 3.8 Flash
- GPT-5 Mini vs Gemini 2.5 Flash
- GPT-5 Nano vs Gemini 2.5 Flash Lite
- GPT-5.4 Mini vs Claude Haiku 4.5