Tested 7 d.Revised Apr. 26
Tool review · Detailed verdict
xAI
Grok 4.20 Heavy
« Fast, plugged into X in real time, leading Heavy agents. Weak spots: blurry GDPR, average French writing. »
Overall score
weighted average · 5 axes
§ Presentation video · Grok 4.20 Heavy
YouTube · officiel
§ Product preview
Price / month
30 €
Pro plan
Cost / 1M tokens
3 €
output rate
Context
256k
max tokens
GDPR
5.5/10
to verify
Hosting
🇺🇸 US
by default
§3.1Strength profile
Scale 0 to 10 · RadarOnAI internal method
§3.2+ Strong points
- 0116-agent Heavy for long reasoning
- 02Real-time X data (live monitoring)
- 03Native multimodal text/image/video/audio
§3.3− Weak points
- 01US hosting, no EU option
- 02French writing less refined than Claude
- 03Unpredictable moderation
Best for
Real-time monitoring, feed analysis, fast comms.
Not for
Polished editorial writing, public sector, sensitive missions.
§04Benchmarks — Grok 4.20 Heavy
Against the rest, cold.
BenchmarkScoreValueRank
GPQA Diamond
Scientific reasoning
79.5%
#3
SWE-bench Verified
Real-world bug fixing
64%
#4
MMLU-Pro
General knowledge
82.5%
#4
HumanEval+
Code generation
87%
#5
FR-Check 2026
Internal language test
71/100
#7
Overall score evolution · Q1 '25 → Q3 '26
Q1 '25Q2 '25Q3 '25Q4 '25Q1 '26Q2 '26Q3 '26
+43
pts in 7 quarters
§ Editor's view · try it?
If your long-form text matters more than your images, Grok 4.20 Heavy remains the subscription that pays back its 30 € monthly fee fastest. We use it daily in the newsroom.
Affiliate link ↗ — RadarOnAI earns a commission if you subscribe, at no extra cost to you. Our verdicts never depend on commissions.
Essayer Grok ↗
No commitment · cancel in one click