xAI releases Grok 4.3 reasoning model with 1M token context at $1.25/M input tokens
xAI has released Grok 4.3, a reasoning model with a 1 million token context window and no output token limit. The model accepts text and image inputs, has always-on reasoning that cannot be disabled, and uses tiered pricing starting at $1.25 per million input tokens and $2.50 per million output tokens.
Grok 4.3 — Quick Specs
xAI releases Grok 4.3 reasoning model with 1M token context at $1.25/M input tokens
xAI has released Grok 4.3, a multimodal reasoning model with a 1 million token context window and no output token limit. Released on April 30, 2026, the model is now available through OpenRouter.
Specifications and pricing
Grok 4.3 processes text and image inputs with text output. Input tokens are priced at $1.25 per million tokens, while output tokens cost $2.50 per million tokens. According to xAI, requests exceeding 200,000 total tokens are billed at a higher rate, though the elevated pricing tier has not been disclosed.
The model features always-on reasoning that cannot be disabled or configured by effort level. This distinguishes it from other reasoning models that allow users to adjust computational intensity.
Technical capabilities
xAI positions Grok 4.3 for agentic workflows, instruction-following tasks, and applications requiring high factual accuracy. The absence of an output token limit, combined with the 1 million token context window, enables the model to handle long-document analysis and multi-step agentic tasks without truncation.
The model supports multimodal input, accepting both text and images, but outputs text only.
API access
Grok 4.3 is accessible through OpenRouter's API, which normalizes requests and responses across providers. The platform routes requests to available providers and includes fallback mechanisms for uptime.
OpenRouter's API supports accessing the model's reasoning process through a reasoning_details array in responses. The platform requires preserving complete reasoning details when passing messages back to the model for continued conversations.
What this means
Grok 4.3 enters a competitive reasoning model market where always-on reasoning represents a trade-off: consistent step-by-step thinking for all queries, but no ability to reduce computational cost for simpler tasks. The 1M token context window and unlimited output position it for enterprise document analysis and complex multi-turn interactions. The tiered pricing structure for requests over 200K tokens suggests xAI expects the model to be used for extended contexts, though the lack of disclosed upper-tier pricing creates uncertainty for budget planning at scale.
Related Articles
Microsoft Releases Mage-VL, a 4B-Parameter Codec-Native Streaming Vision-Language Model
Microsoft has released Mage-VL, a codec-native multimodal foundation model built on a from-scratch 4B-parameter visual encoder paired with Qwen3-4B-Instruct-2507. The model claims up to 3.5x inference speedup over uniform frame sampling and outperforms Qwen3-VL-4B on video and temporal-grounding benchmarks, according to Microsoft.
Unsloth Releases GGUF Quantizations of Kimi K3, a 2.8T-Parameter Open-Weight MoE Model
Unsloth has released GGUF quantizations of Kimi K3, a 2.8-trillion-parameter open-weight Mixture-of-Experts model from Moonshot AI with a 1-million-token context window and native vision support. The largest lossless quantization (Q8) weighs in at 1.56TB.
Alibaba Launches Qwen3.7 Flash: 1M-Context Vision-Language Model at $0.03/$0.13 per 1M Tokens
Alibaba has released Qwen3.7 Flash, a vision-language reasoning model with a 1 million token context window aimed at multimodal agents, visual coding, and computer-use tasks. The model is priced at $0.03 per 1M input tokens and $0.13 per 1M output tokens and is available through OpenRouter.
Microsoft Launches MAI-Cyber-1-Flash, Its First Cybersecurity Model, With Agentic Security Platform Perception
Microsoft has launched MAI-Cyber-1-Flash, its first cybersecurity-specialized model, alongside Perception, an agentic platform for automated threat detection and remediation. The company claims the model outperforms rivals from Anthropic, Google and OpenAI on the Cyber Gym benchmark, though no independent scores have been published.
Comments
Loading...