GPT-5.6 Terra
OpenAI · released
Middle tier of the previous generation, roughly the old “mini” slot. There is no Terra in the GPT-6 lineup, where GPT-6.1 Sol now costs less per output token and scores far higher — so it mostly matters for existing deployments.
Best for: Existing GPT-5.6 deployments · Baseline comparisons
- Intelligence Index
- 42.1
- #15 of 38 · at Max effort
- Price per 1M tokens
- $2 in $12 out
- Blended $4.5 at 3:1
- Context window
- 1.05M
- Up to 128K output
- Output speed
- 108 tokens/s
- $1.4 measured cost per task
Capability profile
One 0–100 score per area, placing this model between the weakest and strongest tracked model across that area's benchmarks.
- Overall66 · #16 of 38
Based on 3 of 3 benchmarks
- Coding60 · #13 of 39
Based on 5 of 5 benchmarks
- Agents & knowledge work64 · #19 of 40
Based on 6 of 6 benchmarks
- Reasoning & science79 · #8 of 39
Based on 2 of 2 benchmarks
- Knowledge & reliability65 · #20 of 39
Based on 2 of 2 benchmarks
- Vision74 · #12 of 31
Based on 2 of 2 benchmarks
Reasoning effort trade-off
The same model at each effort setting. Lower settings give up a few points for a fraction of the cost and wait.
| Setting | Intelligence Index | Cost / task | First answer |
|---|---|---|---|
| Max | 42.1 | $1.4 | 2.3 min |
| xHighSweet spot | 38.0 | $0.63 | 35s |
| High | 34.2 | $0.34 | 3.9s |
| Medium | 30.1 | $0.18 | 1.8s |
| Low | 27.5 | $0.14 | 1.5s |
| No reasoning | 20.8 | $0.14 | 0.8s |
Measured by Artificial Analysis. First answer is the median wait until the first answer token, including thinking time. “Sweet spot” is the cheapest setting that keeps at least 90% of the top score for no more than 60% of its cost.
Benchmark results
All third-party results on record. Rank is among the tracked models that have been tested on that benchmark.
Overall
#15 of 38· Max effort· Artificial Analysis
- Vals Index53.1%
#14 of 26· Max effort· $5.78 per test· Vals AI
- Arena · Text1465 ±4
#16 of 33· xHigh effort· 37,663 votes· Arena
Coding
- Terminal-Bench 4.035.4%
#9 of 38· Max effort· Artificial Analysis
- SciCode55.0%
#14 of 38· Max effort· Artificial Analysis
- Arena · WebDev1522 ±7
#24 of 34· 12,128 votes· Arena
- Vibe Code Bench74.6%
#20 of 26· Max effort· Vals AI
- Code Migration48.3%
#8 of 26· Max effort· Vals AI
Agents & knowledge work
- GDPval-AA47.7
#19 of 39· Max effort· Artificial Analysis
- Arena · Agent+0.4 ±1.2
#19 of 29· XHigh effort· Arena
- Finance Agent v254.4%
#12 of 26· Max effort· Vals AI
#12 of 26· Max effort· Vals AI
- Legal Research Bench41.4%
#13 of 26· Max effort· Vals AI
- Tax Agent Bench65.2%
#11 of 26· Max effort· Vals AI
Reasoning & science
- Humanity's Last Exam42.9%
#16 of 39· Max effort· Artificial Analysis
- CritPt30.0%
#5 of 39· Max effort· Artificial Analysis
- ARC-AGI-2saturating83.9%
#6 of 16· Max effort· $1.09 per task· ARC Prize
- ARC-AGI-30.8%
#5 of 6· Max effort· ARC Prize
- GPQA Diamondsaturating92.5%
#10 of 27· Max effort· Artificial Analysis
Knowledge & reliability
#23 of 39· Max effort· Artificial Analysis
- AA-LCR83.0%
#9 of 39· Max effort· Artificial Analysis
Vision
- MMMU-Pro80.7%
#8 of 27· Max effort· Artificial Analysis
- Arena · Vision1268 ±8
#9 of 19· xHigh effort· 10,292 votes· Arena
Reported by OpenAI
Numbers the lab published at launch, run on its own harness and settings. Useful context, but not independently verified and never used in the rankings on this site.
- SWE-bench Pro · launch table63.4%
- DeepSWE v1.1 · launch table69.6%
- OSWorld 2.0 · launch table50.2%
- GPQA Diamond · launch table92.9%
Specifications
- Lab
- OpenAI
- Released
- Jul 9, 2026
- Availability
- Generally available
- Context window
- 1,050,000 tokens
- Max output
- 128,000 tokens
- Accepts
- Text, Images, PDFs
- Knowledge cutoff
- Feb 2026
- Weights
- Closed
- Reasoning control
- Effort none to max · pro mode
- API model ID
gpt-5.6-terra
Pricing
- Input / 1M tokens
- $2
- Cached input / 1M
- $0.20
- Output / 1M tokens
- $12
- Blended (3:1)
- $4.5
- Measured cost / task
- $1.4
Cut from $2.50 / $15 at the end of July. Batch and Flex are half price.
Official sources
Closest alternatives
The models nearest to this one on overall score.
23 benchmarks tracked. See methodology and sources.