GPT-6.1 Terra Pricing: Full Token Rates and Cost Comparison

GPT-6.1 Terra lists at exactly half the price of GPT-6.1 Sol across the input, cached-input and output rates, and the 30-minute cache write, charging $1.00 per million input tokens, $0.05 per million cached input tokens, and $5.00 per million output tokens compared to Sol's $2.00, $0.10, and $10.00. With the same input/output/cache token mix and pricing modifiers, Terra is cheaper per finished task unless a given workload requires more than twice Sol's total tokens to produce an acceptable result.

At Layer3Labs, we build custom agents and enterprise integrations for businesses choosing model architectures for automated workflows.

OpenAI released GPT-6.1 Terra in late October 2026 as an efficient model for visual reasoning, precise computer use, and natural writing. Whether Terra's output quality matches your production requirements cannot be established by price alone and must be verified directly on your company's tasks.

GPT-6.1 Terra vs Sol Token Rates

Published API rates for GPT-6.1 Terra show a consistent 50 percent reduction across the input, cached-input and output rates when compared to GPT-6.1 Sol. Both models share an identical 1,050,000-token context window with a maximum of 922,000 input tokens and 128,000 output tokens.

Prompt caching reduces recurring input expenses on both models. On Terra, reading cached input costs $0.05 per million tokens, which is exactly 5 percent of the uncached rate, while Sol charges $0.10 per million tokens for cached input. Cache writes on Terra cost $1.25 per million tokens for 30-minute retention and $2.50 for 12-hour retention, while OpenAI’s pricing table lists Sol at $2.50 for 30-minute retention and $5.00 for 12-hour retention.

The baseline rates published by OpenAI for text processing on standard tiers appear below:

• Uncached input per 1M tokens: Terra is $1.00; Sol is $2.00.
• Cached input per 1M tokens: Terra is $0.05; Sol is $0.10.
• Cache write per 1M tokens: Terra is $1.25 (30-minute) or $2.50 (12-hour); Sol is $2.50 (30-minute) or $5.00 (12-hour).
• Output generation per 1M tokens: Terra is $5.00; Sol is $10.00.
• Context window: Both models support 1,050,000 total tokens with 922,000 input and 128,000 output maximums.

What the Same Job Costs on Each Model

Our internal cost analysis, calculated from OpenAI's published rates in the table above, shows identical token volumes cost half as much on Terra as on Sol. To reproduce it: cost = fresh input x input rate + cached input x cached rate + output x output rate. Scenario E applies 2x to input and cache rates and 1.5x to output; C applies x0.5 and D applies x2. We also track broader platform trends in our AI model pricing guide.

For a standard business workload processing 10 million uncached input tokens and 2 million output tokens, Terra costs $20.00 while Sol costs $40.00. If prompt caching satisfies 80 percent of those input tokens, Terra's total drops to $12.40 while Sol drops to $24.80. In an output-heavy job processing 1 million input tokens and 5 million output tokens, Terra totals $26.00 compared to Sol's $52.00.

These token-only estimates assume the stated cache reads hit. They exclude cache writes, cache warmup, tool fees, retries, and taxes. Failed cache reads increase cost.

These scenarios use OpenAI's published text rates. A and B are Standard tier; C, D and E apply the Batch, Fast and long-context modifiers described below:

• Scenario A (10M input, 2M output, no cache): Terra costs $20.00; Sol costs $40.00.
• Scenario B (10M input, 2M output, 80% cached): Terra costs $12.40; Sol costs $24.80.
• Scenario C (10M input, 2M output on Batch): Terra costs $10.00; Sol costs $20.00.
• Scenario D (10M input, 2M output in Fast mode): Terra costs $40.00; Sol costs $80.00.
• Scenario E (100 requests of 300K input and 5K output): Terra costs $63.75; Sol costs $127.50 under long-context multipliers.
• Scenario F (1M input, 5M output): Terra costs $26.00; Sol costs $52.00.

The Break-Even Rule and Testing on Production Tasks

With the same input/output/cache token mix and pricing modifiers, Terra costs less than Sol unless a pipeline consumes more than twice the token volume to deliver an acceptable outcome. Because Terra's rates sit at exactly 0.5 of Sol's rates, running twice as many tokens on Scenario A brings Terra's total to $40.00, which matches Sol exactly. If a task requires 2.5 times the tokens due to repeated retries or verbose reasoning paths, Terra costs $50.00, making Sol the cheaper option at $40.00.

A modest increase in token usage does not eliminate the margin. For example, with the same token mix and modifiers, if Terra requires an extra retry on one out of every two tasks, resulting in an average volume of 1.5 times Sol's tokens, Terra still costs only 75 percent of Sol. You can explore how Sol handles complex reasoning in our GPT-6.1 Sol pricing guide.

Evaluating whether Terra meets your operational standards requires running a structured pilot before routing live production traffic. Teams should run 20 to 50 real business tasks across both models, grade the outputs against a written acceptance standard, and record total tokens per finished task including all failed attempts and retries.

Pricing Modifiers, Rate Limits, and Long Prompts

Both models enforce higher rates when requests exceed specific context thresholds. For any prompt containing more than 272,000 input tokens, OpenAI doubles the input and cache rates and applies a 1.5 multiplier to output rates for the entire request. In Scenario E, 100 requests carrying 300,000 input tokens and 5,000 output tokens bill at $63.75 on Terra and $127.50 on Sol.

Both Terra and Sol support Batch and Flex processing at a 50 percent discount compared to Standard rates, while Fast mode doubles Standard fees. Regional processing carries an optional 10 percent surcharge where supported, while tool calling features such as search and computer use carry per-call fees that are not listed on the model page. Confirm the exact tier and rates on the GPT-6.1 Sol page and the OpenAI pricing page.

Terra operates with defined rate limits across three Standard account tiers, which increase automatically based on usage and spend:

• Build Tier: 5,000 requests per minute and 1,000,000 tokens per minute.
• Launch Tier: 10,000 requests per minute and 4,000,000 tokens per minute.
• Grow Tier: 15,000 requests per minute and 40,000,000 tokens per minute.

Technical Capabilities and Choosing Between Models

OpenAI lists these specifications for each model. Terra has a knowledge cutoff of June 1, 2026, accepts text and image inputs with text-only outputs, and supports reasoning effort settings of none, low, medium (the default), high, xhigh, and max. It operates through Chat Completions, Responses, and Batch APIs, with tool calling restricted to the Responses API.

Sol carries an earlier knowledge cutoff of April 30, 2026, and accepts reasoning effort levels of low, medium (default), high, xhigh, and max. OpenAI recommends Sol for complex coding, computer use, and professional work, whereas it highlights Terra for visual reasoning, precise computer use, and natural writing. For deeper capability breakdowns, review our article on GPT-6.1 Sol explained.

For workloads where an internal pilot confirms that Terra meets your acceptance threshold, default to Terra to capture the 50 percent rate reduction. Keep Sol assigned to complex tasks where Terra fails quality checks, because a failed task that must be rerun can erase your cost savings. This recommendation flips if a workload forces Terra to consume more than 2.0 times Sol's tokens with the same token mix and pricing modifiers, or if OpenAI alters published token rates, so test your specific task prompts directly. The sources do not tell us whether Terra's quality is enough for your tasks; only a pilot answers that.