302.AI
CoverageRelease Notes
OpenAI releases GPT-5.4 Mini and Nano — Weekly AI Newsletter (March 23th 2026) Also: Xiaomi enters the LLM race with MiMo-V2-Pro 😎 News From The Web OpenAI Releases GPT-5.4 Mini and Nano. OpenAI …
$10 off the fastest Kimi K3 and GLM 5.3 from Synthetic
Model details
gpt-5.4-mini-2026-03-17
GPT-5.4 Mini is positioned as OpenAI's mid-tier option in the GPT-5 family, designed to balance strong capability with faster responses and lower cost than the full flagship model. Independent reporting from March 2026 confirms that OpenAI released GPT-5.4 Mini alongside a Nano variant, and the variant carries the broader GPT-5 lineage of general-purpose chat assistants tuned for reasoning, vision, and tool use. Its 272,000-token context window paired with a 128,000-token maximum output makes it well-suited to long documents, multi-turn agent workflows, and retrieval-heavy pipelines where the model must keep substantial material in working memory while still producing lengthy completions.
In practical terms, GPT-5.4 Mini is aimed at developers who want flagship-class reasoning and multimodal understanding without paying flagship-class prices for every token. The listing confirms support for function calling, vision input, reasoning, and web search, while flagging that URL context ingestion is not supported, so teams feeding full webpages should plan around that limitation. Its pricing on OpenAI-compatible gateways positions it as a default choice for new low-latency, high-volume production builds, with the previous GPT-5 mini remaining relevant mainly for legacy workloads where the modest price premium is not justified.
302.AIgpt-5.4-mini-2026-03-17gpt-mini
Quick Info
Powered by- Provider
- 302.AI
- Model key
- gpt-5.4-mini-2026-03-17
- Release date
- Mar 19, 2026
- Last updated
- Mar 19, 2026
- Knowledge cutoff
- 2025-08-31
- Input modalities
- Output modalities
- Capabilities
Cost
- Input token cost
- $0.75
- Output token cost
- $4.50
Limits
- Input tokens
- 272,000 tokens
- Output tokens
- 128,000 tokens
- Context window
- 400,000 tokens
Latest news about gpt-5.4-mini-2026-03-17
302.AI
CoverageBenchmark
GPT-5.4 mini costs more than GPT-5 mini, but OpenAI now recommends it for most new low-latency, high-volume builds. This guide compares price, tool support, benchmarks, and when the cheaper legacy model still makes sense.
302.AI
Coverage
The PolyU ITS eNewsletter explicitly names "GPT-5.4-mini (2026-03-17)" and reports that OpenAI released the model on 17 March 2026, with improvements over GPT-5 mini in coding, reasoning, multimodal understanding, and tool use, running more than 2x faster, and approaching the larger GPT-5.4's performance on SWE-Bench P The PolyU GenAI app deployed this exact version on 1 May 2026, hosted on Microsoft Azure Cloud (not 302.AI), with credit accounting of 0.75 credits per 1K input tokens and 4.5 credits per 1K output tokens including chain-of-thought reasoning. Staff and students can control inference depth via a "reasoning effort" param
302.AI
CoverageBenchmark
The Terminal-Bench 2.0 leaderboard on BenchmarkList explicitly lists "GPT 5.4 Mini 2026-03-17" (high effort, OpenAI) at rank 28 with a score of 44.944% (standard error 4.49), latency of 2241.56, and cost per test of $0.914, with results imported on 2026-07-28. This is an aggregated third-party benchmark, not a self-rep The relatively low ranking (28th out of 30+ models visible) and high latency of 2241.56 alongside a per-test cost of $0.914 suggest that GPT-5.4 Mini 2026-03-17 trails top-tier coding-agent models on this benchmark while being comparatively throughput-constrained and expensive to evaluate. Developers comparing coding-a