OpenAI's GPT-5.6 resets the price-per-token race for India
OpenAI's GPT-5.6 arrives in three tiers from $1 per million tokens, chasing performance-per-dollar. For India's builders, the cheapest model may matter most of all.
The News
OpenAI has released GPT-5.6, its new frontier model family, rolling it out globally on launch day with full availability promised within 24 hours. Rather than a single model, the release ships in three tiers: Sol, the flagship; Terra, pitched at everyday work; and Luna, the most cost-efficient option.
The headline is as much about cost as capability. Sol is priced at $5 per million input tokens and $30 per million output tokens. Terra sits at $2.50 and $15, while Luna undercuts both at $1 input and $6 output. OpenAI's framing is blunt: stronger results for the same spend.
The benchmark sheet is dense. On the Agents' Last Exam evaluation, Sol scores 52.7 per cent, which OpenAI says exceeds rival Claude Fable 5 by 13.1 points. On the Artificial Analysis Coding Agent Index, Sol reaches 80, and on Terminal-Bench 2.1 it hits 88.8 per cent, rising to 91.9 per cent in an Ultra configuration that coordinates four parallel agents by default. Security gains are steep too: ExploitBench climbs to 73.5 per cent from GPT-5.5's 47.9 per cent, and SEC-Bench Pro reaches 71.2 per cent against a prior 45.8 per cent.
OpenAI says it ran roughly 700,000 GPU hours of automated red teaming before launch. Sol is available inside ChatGPT for Plus, Pro, Business and Enterprise users, while developers can reach all three tiers through the API.
Why It Matters
The interesting shift here is not the top score but the price floor. For two years the model race has been fought on raw capability, with cost treated as a footnote. GPT-5.6 inverts that, leading with performance-per-dollar and putting a $1 input tier at the centre of the pitch.
When GPT-4 launched in March 2023, the story was pure novelty; enterprises paid a premium simply to be early. That premium has now largely evaporated. With three competing frontier families trading benchmark leads within a handful of points, the decisive variable becomes unit economics, and OpenAI clearly knows it. Naming a cheap tier Luna and quoting its cost so prominently is a deliberate signal to buyers who count every token.
The security jump is the second signal. Doubling ExploitBench scores in one generation cuts both ways: better defensive tooling for security teams, but also more capable automation available to anyone. Expect that duality to shape the regulatory conversation through the year.
Indian Angle
For Indian developers and startups, the Luna tier is the story. Rupee-denominated budgets have long made frontier models a luxury reserved for funded firms; a $1 input price brings serious capability within reach of bootstrapped teams in Bengaluru and Pune who bill in INR but pay in dollars. Performance-per-dollar is not a marketing line here, it is the difference between shipping and shelving a feature.
It also sharpens the question facing India's own model builders. Sarvam and Krutrim have staked their case on sovereign, cost-aware models tuned for Indian languages and infrastructure. A sub-dollar OpenAI tier compresses the price umbrella they were sheltering under, forcing the value proposition back onto data locality, Indic-language depth and compliance rather than cost alone.
For the large IT services players and MeitY-watching enterprises, cheaper agentic capability accelerates the shift from pilots to production. But the security benchmarks will draw scrutiny too: CERT-In and RBI-regulated entities will want assurances before agents that can coordinate four parallel workflows touch financial systems.
FAQ
When does GPT-5.6 become available?
OpenAI began the global rollout on launch day, with full availability expected within about 24 hours across ChatGPT tiers and the API.
How much does the cheapest tier cost?
Luna is priced at $1 per million input tokens and $6 per million output tokens, making it the most affordable option in the GPT-5.6 family.
What does this mean for Indian startups?
The low Luna pricing lowers the dollar cost of building AI features, easing rupee-budget pressure while intensifying competition for domestic model makers like Sarvam and Krutrim.
Where can I read the original announcement?
OpenAI published the full launch details on its official website, linked in the attribution below.
This story was reported by OpenAI. Read the full original coverage at OpenAI.