OpenAI Launches GPT-6 Sol Pro, a High-Reasoning Mode for Its Mid-Tier Model at $2/$10 per 1M Tokens

OpenAI has added GPT-6 Sol Pro to its model lineup, a configuration of the existing GPT-6 Sol model that runs with reasoning.mode set to pro for higher-quality output on complex tasks. It became available on September 22, 2026, and is listed on OpenRouter with a 1.1 million token context window.

Pricing and performance

GPT-6 Sol Pro is priced at $2 per 1 million input tokens and $10 per 1 million output tokens through OpenAI's standard endpoint, with cached input priced at $0.20 per 1M tokens. According to OpenRouter data, the standard endpoint returns a median (P50) latency of 3.38 seconds and throughput of 35 tokens per second, with 100% uptime recorded over the measurement window.

OpenAI also offers this model through two alternative endpoint tiers: a "Flex" option at $1/$5 per 1M input/output tokens with a P50 latency of 16.70 seconds but higher throughput (176 tokens per second), and a "Fast" tier priced at $4/$20 per 1M tokens, for which detailed latency and throughput figures were not provided. Overall availability across OpenRouter's monitored providers stood at 96.73% over the past 24 hours.

What GPT-6 Sol Pro actually is

According to OpenAI's documentation, GPT-6 Sol Pro is not a separately trained model. It is the same underlying weights as GPT-6 Sol, OpenAI's cost-efficient high-end model that sits below the flagship GPT-6 Astra and above the faster, cheaper GPT-6 Luna in the company's three-tier GPT-6 family. Setting reasoning mode to "pro" changes how the model allocates inference-time compute toward reasoning, intended to improve response quality on demanding tasks without changing the base model itself.

GPT-6 Sol, the underlying model, is described by OpenAI as suited to agentic coding, long-horizon software engineering in real codebases, business workflow automation, and computer-use tasks, with factual reliability and communication style OpenAI says approaches its flagship GPT-6 Astra at a lower cost. GPT-6 Sol itself is priced separately at $2/$10 per 1M tokens (standard) or $1/$5 per 1M tokens on a lower-cost tier, matching the Pro variant's standard pricing — since it is the same model, only the reasoning configuration differs.

The broader GPT-6 family available on OpenRouter includes GPT-6 Astra (flagship, $10/$50 or $5/$25 per 1M tokens) and GPT-6 Luna (fast tier, $0.10/$0.50 or $0.05/$0.25 per 1M tokens), each also offered in a "Pro" reasoning-mode variant at matching prices to their base models. All GPT-6 tier models share the 1.1 million token context window.

What this means

GPT-6 Sol Pro is not a new model release in the traditional sense — it is a reasoning-mode configuration of an already-released model, GPT-6 Sol. OpenAI's decision to list Pro variants as distinct SKUs on OpenRouter, each with its own pricing and performance profile, gives developers granular control over the latency/cost/quality tradeoff without needing to manage separate model identifiers for weights that are identical. For teams building agentic or coding workflows, the practical decision is whether the added reasoning depth justifies the same per-token price as standard Sol, given that quality gains here are reasoning-effort driven rather than architectural.