OpenAI launches GPT-6.1 Terra: Mid-tier AI model matches Astra at one-tenth the cost

OpenAI introduces GPT-6.1 Terra as cost-efficient mid-tier AI model

OpenAI has launched GPT-6.1 Terra, positioning it as a mid-tier AI model that bridges the gap between premium performance and budget constraints1

1

. The model joins OpenAI's four-tier lineup, slotting between the budget GPT-6 Luna and the premium GPT-6.1 Sol and GPT-6 Astra options. Terra's API pricing stands at $1 per million input tokens and $5 per million output tokens, representing a 50% reduction compared to both GPT-5.6 Terra and GPT-6.1 Sol2

2

. This positions Terra as a cost-effective alternative for developers automating high-volume work such as invoice processing and data entry tasks.

Performance benchmarks reveal competitive capabilities against flagship models

On Agents' Last Exam, which measures AI performance on professional computer tasks, GPT-6.1 Terra scored 15 percentage points higher than GPT-5.6 Terra at medium reasoning effort while costing approximately one-third as much3

3

. More remarkably, Terra matched GPT-6 Astra's performance on the same benchmarks at one-tenth the price per token5

5

. The model demonstrated substantial improvements in visual reasoning, scoring competitively on MMMU-Pro against both Astra and GPT-6.1 Sol. For coding tasks, Terra achieved a 50.3% score on Terminal-Bench 4.0 at maximum reasoning effort, more than double GPT-5.6 Terra's 21.5% performance2

2

.

Speed advantages position Terra for everyday computer work automation

GPT-6.1 Terra completes computer-use tasks more than three times faster than both GPT-6 Astra and GPT-6.1 Sol4

4

. This speed advantage makes Terra particularly suited for everyday computer work including filling out forms, editing videos, interpreting technical drawings, and automating repetitive desktop tasks. Ted Sanders, a member of technical staff at OpenAI, explained that Terra excels at technical communication, matches higher-tier models on visual reasoning and computer use while operating much faster, though it performs slightly worse on most other tasks1

1

. The model's knowledge cutoff date of June 1, 2026, makes it the most current in OpenAI's model family, surpassing the April 30, 2026 cutoff for Astra and Sol.

Reduced hallucination rates improve reliability for enterprise applications

GPT-6.1 Terra demonstrates significantly lower hallucination rates compared to its predecessor. On OpenAI's internal benchmarks, the error rate at low reasoning effort dropped from 53.6% with GPT-5.6 Terra to 16.7% with GPT-6.1 Terra3

3

. At maximum reasoning effort, the error rate fell by 18 percentage points to 4.3%, nearly matching the rates for GPT-6 Astra and GPT-6.1 Sol2

2

. These improvements make Terra more reliable for Enterprise users deploying the model in production environments where accuracy matters.

Lower pricing challenges assumptions about AI economics

The launch reflects broader trends in AI model economics that defy expectations of rising costs. Data scientist Max Woolf noted that despite widespread predictions that OpenAI and Anthropic would need to raise inference costs to match their infrastructure spending, prices continue declining while model performance improves1

1

. Artificial Analysis observed that GPT-6.1 Terra pushes the cost-efficiency frontier, delivering more intelligence for less than a third of the cost compared to GPT-5.6 Terra. This pricing pressure stems from competitive dynamics where inference providers reduce token costs to compete, forcing OpenAI and Anthropic to lower their own pricing.

Availability spans ChatGPT Work, Codex, and API access for developers

GPT-6.1 Terra is available immediately to ChatGPT Plus, Pro, Enterprise, and Edu users through ChatGPT Work and Codex5

5

. Developers can access the model through the API using the identifier gpt-6.1-terra. Cached input tokens cost $0.05 per million tokens, further reducing expenses for applications with repetitive queries. OpenAI maintains its recommendation that developers use GPT-6.1 Sol or GPT-6 Astra for the most complex coding, reasoning, and research workloads, positioning Terra as the optimal choice for mid-range applications balancing capability and cost. The model reaches the "Critical" level for cybersecurity skills under OpenAI's Preparedness Framework and carries the same safety protections as GPT-6 Astra3

3

.