What is Grok 4.20 Non-Reasoning?
Grok 4.20 Non-Reasoning is xAI's Grok 4.20 with the thinking phase switched off: the same weights and price as Grok 4.20 Reasoning, answering immediately instead of reasoning first. It went live on the API on March 10, 2026 as the 0309 snapshot and got refreshed weights in April 2026. It accepts text and image input, outputs text, and runs on Unifically as xai/grok-4.20-0309-non-reasoning with a 1M-token context window. xAI built the 4.20 line for speed, agentic tool calling, strict prompt adherence, and a low hallucination rate, and this model is the fast end of it.
Key features of Grok 4.20 Non-Reasoning
First token in 0.64 seconds
Artificial Analysis measures 0.64 seconds to first token against a 1.40-second median for non-reasoning models in its class, and 95.8 output tokens per second, #12 of 64. Replies start before a user notices the wait.
Same 1M context and tools as Reasoning
A million tokens of input, image understanding, function calling, and structured outputs, with none of the reasoning tokens on the bill.
Strict prompt adherence
xAI tuned the 4.20 line to follow the instruction as written, which is what you want from a model that answers without a second pass.
Best for
Real-time chat
Sub-second first tokens for assistants and support widgets.
Extraction and classification
Structured JSON from tickets, forms, and documents at high volume.
Routing and moderation
Quick decisions where the label matters more than the reasoning.
Templated generation
Product copy, summaries, and replies that follow a fixed format.
Long-context lookup
Find and quote from a million tokens without a thinking delay.
Visual input tasks
Screenshots, forms, and charts read as input for text answers.
Use cases
Use Grok 4.20 Non-Reasoning where latency is the product: live chat, autocomplete-style suggestions, and voice-adjacent flows that need the first words within a second. It handles high-volume extraction and classification over tickets, forms, and logs with structured outputs, and routes requests to the right tool or model without adding reasoning tokens to every call. With a 1M-token window it can also answer lookup questions over a whole document set. Pair it with Grok 4.20 Reasoning behind the same endpoint and switch model IDs when a request needs thought.
Limitations
No thinking means less intelligence. Artificial Analysis scores it 22 on the Intelligence Index, just under the 23 median for non-reasoning models and far below Grok 4.20 Reasoning at 38. Multi-step problems, math, and tricky code belong on a reasoning model.
Speed is the whole pitch; there are no independent coding or agent board results for this model, and the 0309 weights have already been refreshed once, so re-check outputs after xAI updates.
Grok 4.20 Non-Reasoning vs Reasoning
Same weights, same price, one difference. Non-Reasoning answers in 0.64 seconds to first token at 95.8 tokens per second; Reasoning takes about 16 seconds to start but scores 38 against 22 on the Intelligence Index and 72.2% on SWE-bench Verified. Put Non-Reasoning in front of users and on batch pipelines; put Reasoning behind anything that has to be right the first time.
When to use Grok 4.20 Non-Reasoning
Use it when a request has to start answering instantly, when the task is extraction or routing, or when reasoning tokens would double the cost for no visible gain. Skip it for hard reasoning, coding beyond quick fixes, and research-style questions, which belong on Grok 4.20 Reasoning, Grok 4.5, or Grok 4.6.
API examples
Call Grok 4.20 Non-Reasoning from any language by POSTing to /v1/chat/completions, the OpenAI-compatible endpoint shared by every language model on the platform. Full parameter docs live at docs.unifically.com/models/llm/xai/grok-4.20-0309-non-reasoning.
curl -X POST https://api.unifically.com/v1/chat/completions \
-H "Content-Type: application/json" \
-H "Authorization: Bearer YOUR_API_KEY" \
-d '{
"model": "xai/grok-4.20-0309-non-reasoning",
"messages": [
{ "role": "user", "content": "Classify this support message as billing, bug, or feature request and reply with one word." }
]
}'
The response comes back synchronously with the completion. Set "stream": true to receive tokens as they generate.
FAQs
People also ask
xai/grok-4.20-0309-non-reasoning, called through the OpenAI-compatible POST /v1/chat/completions endpoint with your Unifically API key. The same model ID also works on /v1/responses and /v1/messages.
Text and image input, text output. The context window is 1M tokens, and function calling, structured outputs, and streaming all work through the standard chat completions request.
Artificial Analysis measures 0.64 seconds to first token, against a 1.40-second median for non-reasoning models in its class, and 95.8 output tokens per second,
No. It skips the reasoning phase entirely, which is where the speed comes from. Use Grok 4.20 Reasoning instead if you need thinking.
It scores 22 on the Artificial Analysis Intelligence Index, just under the 23 median for non-reasoning models. Grok 4.20 Reasoning scores 38.
For extraction, classification, routing, templated replies, and any chat where the answer has to start in under a second. Both models share the same price and 1M context window.
