GPT-5.6 Terra benchmarks

GPT-5.6 Terra is a proprietary model from OpenAI, released 9 Jul 2026. It ranks 28th of 51 on our overall leaderboard and 12th of 51 for long context. At A$6.32 per million tokens (blended), it costs 1.7× the median model we track. It writes about 92 tokens a second, and answers start after 2.9 min on average, thinking included. It can be called from AWS Bedrock in Sydney and Melbourne and Microsoft Azure in Australia East, but only through global routing, so requests may be processed outside Australia.

#28 OF 51 OVERALLA$6.32 / 1M TOKENS1.05M CONTEXTUPDATED 23 SEP 2026
CONTEXT WINDOW
1,050,000 tokens
MAX OUTPUT
128,000 tokens
INPUT
text, image
WEIGHTS
Proprietary
RELEASED
9 Jul 2026
OUTPUT SPEED
92 tok/s
ANSWER STARTS AFTER
2.9 min (thinking included)
OPENAI DOCS
§ 02 — RESULTS

Benchmark results.

Independent test scores are for the max effort setting. “Among tracked” ranks GPT-5.6 Terra against the 51 models on this site; LMArena ranks run across its full leaderboard.

ALL RESULTS8 OF 8 CORE SIGNALS
GPT-5.6 Terra benchmark results
BENCHMARKRESULTAMONG TRACKEDSOURCE DETAIL
INDEPENDENT TESTS · ARTIFICIAL ANALYSIS
AA Intelligence Index42.1#16 of 51max effort
AA Coding Index76.7#6 of 42max effort
GPQA Diamond92.5%#18 of 44max effort
Humanity’s Last Exam42.9%#22 of 51max effort
AA-LCR83.0%#12 of 51max effort
Terminal-Bench 2.188.0%#6 of 42max effort
τ²-Bench (banking)40.2%#15 of 41max effort
SciCode55.0%#21 of 46max effort
BLIND HUMAN VOTES · LMARENA
LMArena Text28,119 votes · ±51466#30 of 42#45 of 402 on LMArena · 13 Sep 2026
LMArena WebDev9,984 votes · ±71518#31 of 45#39 of 129 on LMArena · 22 Sep 2026
LMArena Vision7,726 votes · ±81267#23 of 33#32 of 152 on LMArena · 13 Sep 2026
LMArena Document5,787 votes · ±81477#11 of 26#14 of 44 on LMArena · 13 Sep 2026
LMArena Agent20,301 sessions#23#20 of 37#23 of 46 on LMArena · 15 Sep 2026
§ 03 — COST

What GPT-5.6 Terra costs in Australian dollars.

List prices converted at the snapshot’s RBA rate, excluding GST. Reasoning models are billed for their thinking as output tokens, so heavier settings cost more than the job estimates show.

LIST PRICE
GPT-5.6 Terra price per million tokens
PER MILLION TOKENSAUDUSD LIST
Input tokensA$2.81US$2.00
Output tokensA$16.85US$12.00
Blended (3 in : 1 out)A$6.32US$4.50
EVERYDAY JOBSESTIMATES
Estimated cost of common jobs in Australian dollars
JOBCOST (AUD)MEDIAN MODEL
Customer support replyPER 1,000 REPLIESA$10.67A$6.58
Summarise a 30-page documentPER 100 DOCUMENTSA$6.96A$4.74
Agentic coding taskPER 10 TASKSA$5.56A$3.72
SETTINGSMEASURED SEPARATELY
GPT-5.6 Terra settings compared
SETTINGINTELLIGENCEA$ / 1MSPEED
max effort42.1A$6.3292 tok/s
xhigh effort38.0A$6.3280 tok/s
high effort34.2A$6.3279 tok/s
medium effort30.1A$6.3276 tok/s
low effort27.5A$6.3274 tok/s
Non-reasoning20.8A$6.3273 tok/s
§ 04 — AUSTRALIA

Running GPT-5.6 Terra in Australia.

It can be called from AWS Bedrock in Sydney and Melbourne and Microsoft Azure in Australia East, but only through global routing, so requests may be processed outside Australia. How the platforms compare →

AUSTRALIAN CLOUD REGIONS
GPT-5.6 Terra availability in Australian cloud regions
PLATFORMAVAILABILITYDETAIL
AWS Bedrock · SydneyGLOBAL

Provider docs, checked 23 Sep 2026

AWS Bedrock · MelbourneGLOBAL

Provider docs, checked 23 Sep 2026

Azure · Australia EastGLOBAL

Provider docs, checked 23 Sep 2026

Google Vertex AI · SydneyNOT OFFERED
§ 06 — QUESTIONS

GPT-5.6 Terra, answered.

GPT-5.6 Terra lists at US$2.00 per million input tokens and US$12.00 per million output tokens — A$2.81 and A$16.85 at A$1 = US$0.7123 (RBA, 22 Sep 2026), excluding GST. A typical customer support reply works out at about A$10.67 per 1,000 replies, before any reasoning tokens.

GPT-5.6 Terra can be called from AWS Bedrock in Sydney and Melbourne and Microsoft Azure in Australia East, but only through global routing, so requests may be processed outside Australia. Availability comes from the cloud providers’ own documentation; check your provider’s current terms before relying on it for data-residency obligations.

It ranks 14th of 42 on our coding ranking, with an Artificial Analysis Coding Index of 76.7 and an LMArena WebDev rating of 1518 (39th of 129). The current leader is Claude Fable 5.1.

It ranks 16th of 41 on our agents ranking, completing 40.2% of τ²-Bench banking customer-service tasks and 88.0% of Terminal-Bench 2.1 tasks.

Artificial Analysis measures GPT-5.6 Terra at about 92 tok/s of output, with the answer starting after 2.9 min on average once thinking time is included (max effort setting).

OpenAI documents a context window of 1,050,000 tokens (1.05M), with up to 128,000 output tokens. On AA-LCR, which tests reasoning across ~100,000-token document sets, it scores 83.0%.

PUT THE COMPARISON TO WORK

Need help choosing and using AI for your business?