GPT-5 models

9 versions from OpenAI, reviewed by the TopReviewed AI panel.

ModelScoreStatusReleasedPrice in/outContext
GPT-5.5 Pro
frontier deep-research and hardest reasoning
7.7GA2026-04-23$30 / $1801.1MReview →
GPT-5.5
frontier agentic coding and computer use
8.7GA2026-04-22$5 / $301.1MReview →
GPT-5.4 mini
best price-to-capability in the OpenAI lineup
9.0GA2026-03-16$0.75 / $4.50400KReview →
GPT-5.4 nano
cheapest viable frontier-family worker tier
8.2GA2026-03-16$0.20 / $1.25400KReview →
GPT-5.4
default production workhorse
8.7GA2026-03-04$2.50 / $151.1MReview →
GPT-5.4 Pro
escalation tier for in-flight GPT-5.4 stacks
6.7GA2026-03-04$30 / $1801.1MReview →
GPT-5
legacy flagship on a migration path
6.5GA2025-08-06$1.25 / $10400KReview →
GPT-5 nano
OpenAI cost floor for plumbing-grade work
6.4GA2025-08-06$0.05 / $0.40400KReview →
GPT-5 mini
superseded legacy mini, migrate to GPT-5.4 nano
5.8GA2025-08-06$0.25 / $2400KReview →

Strongest: GPT-5.4 mini

GPT-5.4 mini is OpenAI's high-volume workhorse, released 2026-03-17 alongside GPT-5.4 nano — the "most capable small model" OpenAI had shipped, built explicitly for the subagent era. It nearly matches full GPT-5.4 on coding (SWE-bench Pro 54.4% vs base 57.7%) at roughly one-third the cost, with a full tool surface including computer use that is rare at this price. The one-sentence buyer's take: for cost- and latency-sensitive workloads that still need strong reasoning and tool use, this is the default mini-tier choice in the OpenAI lineup as of 2026-05-28. - Provider: OpenAI - Release: 2026-03-17 - Status: GA - Context: 400,000 tokens - Max output: 128,000 tokens - Modalities: text + image in, text out - Knowledge cutoff: 2025-08 - Headline price: $0.75 in / $4.50 out per 1M tokens

Full review →

Best value: GPT-5.4 nano

GPT-5.4 nano is OpenAI's lightweight workhorse, released 2026-03-17 — the cheapest model in the GPT-5.4 frontier family, built for the highest-volume, lowest-latency workloads. Despite being four tiers below the flagship, it beats the previous-generation GPT-5 mini on SWE-bench Pro (52.4% vs 45.7%) and matches it on GPQA Diamond (82.8%). The one-sentence buyer's take: indispensable as the bottom tier of a routed architecture — classification, extraction, triage, and routing at a price that competes with embedding pipelines. - Provider: OpenAI - Release: 2026-03-17 - Status: GA - Context: 400,000 tokens - Max output: 128,000 tokens - Modalities: text + image in, text out - Knowledge cutoff: 2025-08 - Headline price: $0.20 in / $1.25 out per 1M tokens

Full review →