Curiosity · AI Model

Grok 1.5

Grok 1.5 was xAI's first substantial update to its flagship LLM, released in March 2024 to X Premium subscribers and early-access partners. Compared with the open-sourced Grok-1 (314B MoE), Grok 1.5 was a from-scratch retraining focused on reasoning and long-context — extending the context window to 128k tokens and delivering large jumps on MATH, GSM8K, and HumanEval. A vision variant (Grok 1.5V) followed in April 2024, adding real-world reasoning over images.

Model specs

Vendor
xAI
Family
Grok
Released
2024-03
Context window
128,000 tokens
Modalities
text, vision

Strengths

  • Big reasoning jump over Grok-1
  • 128k context was competitive at 2024 release
  • Introduced xAI's vision capability (Grok 1.5V)
  • Powered the Grok chat experience on X during 2024

Limitations

  • Fully superseded — Grok 2, 3, and 4 all beat it
  • Not offered on current xAI API
  • No open weights (unlike Grok-1)
  • Safety tuning weaker than contemporaries

Use cases

  • Historical reference for Grok-series progression
  • X (Twitter) chat during 2024
  • Benchmarking long-context LLMs at 128k
  • Vision variant (Grok 1.5V) — early xAI multimodal pilot

Benchmarks

BenchmarkScoreAs of
MATH≈50%2024-03
GSM8K≈90%2024-03
HumanEval≈74%2024-03

Frequently asked questions

What is Grok 1.5?

Grok 1.5 is xAI's March 2024 update over Grok-1 — a retrained LLM with 128k context and substantially stronger reasoning, math, and code benchmarks.

Is Grok 1.5 open-source?

No — unlike the original Grok-1 (whose weights were released), Grok 1.5 was never open-sourced.

Should I use Grok 1.5 today?

No. Grok 2 and Grok 3/4 supersede it. Grok 1.5 is primarily a historical reference point.

Sources

  1. xAI — Announcing Grok 1.5 — accessed 2026-04-20
  2. xAI — Grok 1.5V vision — accessed 2026-04-20