OpenRouter Adds Grok 4.3 With Massive Agentic Performance Jump and Lower Pricing

May 1, 2026 · Updated May 13, 2026

OpenRouter, a unified API for accessing hundreds of LLMs, added Grok-4.3 from xAI to its catalog. This multimodal reasoning model (an AI that performs step-by-step logic before answering) supports text and image inputs with a 1 million token context window. Reasoning is always active and cannot be disabled.

Context window: 1,000,000 tokens
Pricing (input): $1.25 per million tokens
Pricing (output): $2.50 per million tokens
Benchmark (GDPval-AA): 1500 ELO
Benchmark (Intelligence Index): 53.2
Throughput: 108 tokens per second
Availability: OpenRouter API

The update marks a major shift in agentic performance (AI that autonomously plans and executes multi-step tasks). Grok-4.3 saw a 321-point increase to 1500 ELO on the GDPval-AA benchmark for economically valuable workflows. This mirrors OpenRouter's GPT-5.5 Pro launch, which also prioritized inspectable reasoning for complex agent pipelines.

You can deploy Grok-4.3 for long-document analysis and deep research requiring high factual accuracy. Pricing is $1.25 per million input tokens and $2.50 per million output tokens, with tiered rates for requests exceeding 200,000 tokens. The model is available now via the OpenRouter API and chat interface.

View the full update on openrouter.ai

OpenRouter

@OpenRouterApr 30

The new Grok-4.3 from @xai is live on OpenRouter! Grok-4.3 releases at a lower price than Grok-4.2, while seeing a large jump in agentic performance: a 321 point increase to 1500 ELO on @ArtificialAnlys GDPval-AA, surpassing other top models despite the lower price. https://t.co/Uj3h5cvIB8

2391.9k

View on X

Still wondering? A few quick answers below.

Grok-4.3 is the latest reasoning model from xAI, designed for high factual accuracy and complex agentic workflows. It is a multimodal model that accepts both text and image inputs while producing text outputs. It is specifically optimized for long-document analysis and multi-step tasks that require the model to think through logic before responding.

On OpenRouter, Grok-4.3 is priced at 1.25 dollars per million input tokens and 2.50 dollars per million output tokens for requests under 200,000 tokens. For larger requests exceeding that threshold, the price increases to 2.50 dollars for input and 5.00 dollars for output per million tokens. Cache reads are also available at discounted rates.

Reasoning in Grok-4.3 is a core capability where the model performs internal chain-of-thought processing before delivering a final answer. Unlike some other models, reasoning is always active in Grok-4.3 and cannot be disabled or adjusted by effort level. Developers can access these internal thinking traces through the OpenRouter API to monitor the model's logic.

Grok-4.3 features a massive 1 million token context window, allowing it to process and analyze extremely long documents or large codebases in a single request. While the input context is expansive, the model also supports a 1 million token output limit, making it suitable for generating long-form content or detailed research reports without truncation.

Grok-4.3 delivers a significant jump in agentic performance compared to its predecessor. It achieved a 321-point increase on the Artificial Analysis GDPval-AA benchmark, reaching a 1500 ELO score. This improvement allows it to surpass other top-tier models in executing economically valuable tasks and autonomous workflows, all while being offered at a lower starting price.