GPT-5.4 vs GPT-5.6 Terra
GPT-5.6 Terra leads GPT-5.4 by 7.2 points on the Shogo Agent Index, but their ranges overlap, so treat it as close. GPT-5.4 wins 4 and GPT-5.6 Terra wins 14 of 18 shared benchmarks. GPT-5.6 Terra is cheaper for agent workloads.
Add more models to this comparison →Pillars
| Pillar | GPT-5.4 | GPT-5.6 Terra |
|---|---|---|
| Agentic coding | 57 | 66 |
| Long-horizon autonomy | 61 | 69 |
| Knowledge-work agents | 40 | 47 |
| Computer use | 42 | 48 |
| Security & context learning | 54 | 56 |
Shared benchmarks
| Benchmark | GPT-5.4 | GPT-5.6 Terra |
|---|---|---|
| APEX-Agentsindex | 52.4% | 58.2% |
| DeepSWEindex | 51.8% | 69.6% |
| Vending-Bench 2index | $6,144 | $7,343 |
| ARC-AGI | 93.7% | 96.5% |
| ARC-AGI-2 | 74.0% | 83.9% |
| Chess Puzzles | 48.0% | 47.0% |
| CritPt | 23.4% | 30.0% |
| DTBench | 94.4% | 93.3% |
| FrontierMath Tier 4 | 49.0% | 70.7% |
| FrontierMath Tiers 1-3 | 78.6% | 86.0% |
| GPQA Diamond | 93.3% | 93.3% |
| LMCA | 52.0% | 54.9% |
| MirrorCode | 23.3% | 41.1% |
| OTIS Mock AIME | 97.8% | 99.7% |
| ProofBench | 56.0% | 74.0% |
| SciCode | 56.6% | 55.0% |
| SimpleQA Verified | 45.1% | 43.2% |
| WebDev Arena | 1463 | 1521 |
Specs and price
| GPT-5.4 | GPT-5.6 Terra | |
|---|---|---|
| Agent Index | 51.6 (#33) | 58.8 (#16) |
| Agent $/M | $1.27 | $1.02 |
| Input / output $/M | $2.50 / $15.00 | $2.00 / $12.00 |
| Context | 1.1M | 1.1M |
| Released | Mar 5, 2026 | Jul 9, 2026 |
| Weights | Closed | Closed |
| API hosts | 4 | 4 |
Don’t pick one. Route.
Shogo can run GPT-5.4 and GPT-5.6 Terra side by side and send each step of an agent to the one that’s cheaper for it.
Try both in Shogo