OpenAI: GPT-5.6 Terra

openai/gpt-5.6-terra

GPT-5.6 Terra (OpenAI) ranks #37 of 260 on the TryAii Score at $2/$12 per million input/output tokens, though Ling 3.1 Flash matches it for $0.225 per million output tokens. It's strongest on knowledge (#18 on GPQA). It's reasonably quick at roughly 75-94 tokens/sec.

Prompt $/1M

$2.000

Completion $/1M

$12.000

Context

1,050,000 tok

Best provider

116 tok/s · Artificial Analysis

Specs

Model id: openai/gpt-5.6-terra
Display name: OpenAI: GPT-5.6 Terra
Modality: text+image+file->text
Tokenizer: GPT
Instruct type: —
Context length: 1,050,000
Max completion: 128,000
Prompt $/1M: $2.000
Completion $/1M: $12.000
Free tier: no

Capabilities

Function callingno
JSON modeno
System promptno
Visionno
Moderatedyes

Benchmark scores

Each axis is one benchmark, normalized to the best model in the field.

No benchmark data for this model.

Speed by provider

Provider	tok/s	TTFT	Total latency
Artificial Analysis	116	1,814 ms	—
Azurefastest	87	784 ms	798 ms
Amazon Bedrock	85	1,539 ms	2,018 ms
OpenAI	45	1,204 ms	1,294 ms

Prompt caching by provider

Provider	Min tokens to cache	Notes
OpenAI	1,024 tokens	GPT-5.6 family — first OpenAI generation with a cache-write premium (1.25x input on cold writes; single-use large prompts pay it for nothing). Docs cite prefixes staying reusable at least 30 min.
Amazon Bedrock	1,024 tokens	GPT-5.6 on Bedrock: explicit cachePoint (unlike OpenAI direct where it is automatic); 30-min TTL; 90% read discount, 1.25x write premium.