Grok 4.20 Multi-Agent vs Grok 4.7
Grok 4.20 Multi-Agent costs 2.4x less per output token than Grok 4.7 ($2.50 vs $6.00 per million). Updated October 21, 2026; no benchmark has tested both models yet.
- Coding
- No data
- Overall
- No data
- Output $/M
- $2.50
- Coding
- 66.7
- Overall
- 46.4
- Output $/M
- $6.00
Grok 4.20 Multi-Agent or Grok 4.7: which to pick
Only Grok 4.7 has a coding score, 66.7 on the Vibe Coding Index; Grok 4.20 Multi-Agent has none, so the two cannot be ranked on coding.
Pick Grok 4.20 Multi-Agent for output tokens 2.4x cheaper ($2.50 vs $6.00 per million) and a larger context window (2M against 500K tokens).
No benchmark has tested both yet. Grok 4.20 Multi-Agent has none yet; Grok 4.7 on LiveBench coding, LMArena WebDev and Terminal-Bench 4 (Vals AI). Compare them on price and context instead.
Which one for your work
Each card is decided by one number both models carry. Where only one has it, the card says so and makes no call.
Coding
Vibe Coding Index, six public coding benchmarks
No call
Only Grok 4.7 has a coding score, so no call is made.
One side not measured
Intelligence
AA Intelligence Index
No call
Only Grok 4.7 has a AA Intelligence Index result, 46.4.
Directional only
Web design
Design Arena: websites
No call
Only Grok 4.7 has a Design Arena: websites result, 1254.
Directional only
General
Vals Index
No call
Only Grok 4.7 has a Vals Index result, 58.6%.
Directional only
Long documents
Listed context window
Grok 4.20 Multi-Agent
Grok 4.20 Multi-Agent takes 2M tokens against 500K.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
Grok 4.20 Multi-Agent
Grok 4.20 Multi-Agent $0.0025, Grok 4.7 $0.0050 per turn.
List prices
At a glance
- Coding index
- Grok 4.20 Multi-Agent not benchmarked, Grok 4.7 66.7 from 5 benchmarks.
- Widest category
- None yet: no category has been tested on both models.
- Widest single gap
- None yet: no benchmark has tested both models.
- Output price
- Grok 4.20 Multi-Agent $2.50 against Grok 4.7 $6.00 per million output tokens, 2.4x apart.
- Context window
- Grok 4.20 Multi-Agent takes 2M tokens, Grok 4.7 500K.
Benchmark by benchmark
Every board either model has a result on, a category to a fold; bars are each result on a 0 to 100 scale. Only one model is measured on each, so no row names a leader.
CodingDirectional onlySix of these boards feed the Vibe Coding Index18 boards measured
| Benchmark | Grok 4.20 Multi-Agent | Grok 4.7 | Leader |
|---|---|---|---|
| LiveBench coding | Not measured | 84.5% | Only Grok 4.7 measured |
| LMArena WebDev | Not measured | 1637 | Only Grok 4.7 measured |
| Terminal-Bench 4 (Vals AI) | Not measured | 28.8% | Only Grok 4.7 measured |
| Vibe Code Bench (Vals AI) | Not measured | 86.2% | Only Grok 4.7 measured |
| Code migration (Vals AI) | Not measured | 44.8% | Only Grok 4.7 measured |
| ProgramBench (Vals AI) | Not measured | 0.5% | Only Grok 4.7 measured |
| IOI (Vals AI) | Not measured | 57.7% | Only Grok 4.7 measured |
| Terminal-Bench 2.1 (Vals AI) | Not measured | 73.4% | Only Grok 4.7 measured |
| SWE-bench Multimodal (Vals AI) | Not measured | 42.9% | Only Grok 4.7 measured |
| CyberBench (Vals AI) | Not measured | 69.5% | Only Grok 4.7 measured |
| SciCode | Not measured | 57.8% | Only Grok 4.7 measured |
| FrontierCode | Not measured | 47.6% | Only Grok 4.7 measured |
| FrontierSWE | Not measured | 29.5% | Only Grok 4.7 measured |
| LMArena Text: coding | Not measured | 1432 | Only Grok 4.7 measured |
| LMArena Text: software and IT | Not measured | 1421 | Only Grok 4.7 measured |
| LMArena WebDev: React | Not measured | 1663 | Only Grok 4.7 measured |
| LMArena WebDev: HTML | Not measured | 1627 | Only Grok 4.7 measured |
| LMArena WebDev: from an image | Not measured | 1586 | Only Grok 4.7 measured |
| Neither model has a result on the other 20 boards here | |||
IntelligenceDirectional onlyPublic boards, shown as published, kept out of the index2 boards measured
| Benchmark | Grok 4.20 Multi-Agent | Grok 4.7 | Leader |
|---|---|---|---|
| AA Intelligence Index | Not measured | 46.4 | Only Grok 4.7 measured |
| Epoch Capabilities Index | Not measured | 154.4 | Only Grok 4.7 measured |
Agentic workDirectional onlyPublic boards, shown as published, kept out of the index15 boards measured
| Benchmark | Grok 4.20 Multi-Agent | Grok 4.7 | Leader |
|---|---|---|---|
| LiveBench research | Not measured | 86.4% | Only Grok 4.7 measured |
| LiveBench browser | Not measured | 60.0% | Only Grok 4.7 measured |
| LiveBench long agentic | Not measured | 50.1% | Only Grok 4.7 measured |
| SREBench (Vals AI) | Not measured | 3.8% | Only Grok 4.7 measured |
| Finance Agent (Vals AI) | Not measured | 52.3% | Only Grok 4.7 measured |
| Tax Agent Bench (Vals AI) | Not measured | 34.2% | Only Grok 4.7 measured |
| Legal Research Bench (Vals AI) | Not measured | 47.1% | Only Grok 4.7 measured |
| Harvey's Legal Agent Benchmark (Vals AI) | Not measured | 18.3% | Only Grok 4.7 measured |
| Excel Modeling Benchmark (Vals AI) | Not measured | 67.0% | Only Grok 4.7 measured |
| BioMysteryBench (Vals AI) | Not measured | 69.3% | Only Grok 4.7 measured |
| Terminal-Bench-Science (Vals AI) | Not measured | 10.0% | Only Grok 4.7 measured |
| Vals Science Index | Not measured | 34.8% | Only Grok 4.7 measured |
| Robo-bench (Vals AI) | Not measured | 0.2% | Only Grok 4.7 measured |
| APEX-Agents | Not measured | 54.6% | Only Grok 4.7 measured |
| Vending-Bench 2 | Not measured | $10,537 | Only Grok 4.7 measured |
| Neither model has a result on the other 24 boards here | |||
ReasoningDirectional onlyPublic boards, shown as published, kept out of the index13 boards measured
| Benchmark | Grok 4.20 Multi-Agent | Grok 4.7 | Leader |
|---|---|---|---|
| LiveBench reasoning | Not measured | 82.7% | Only Grok 4.7 measured |
| MysteryMechanism (Vals AI) | Not measured | 25.2% | Only Grok 4.7 measured |
| Humanity's Last Exam (Artificial Analysis run) | Not measured | 43.1% | Only Grok 4.7 measured |
| GPQA Diamond (Epoch AI run) | Not measured | 92.7% | Only Grok 4.7 measured |
| ARC-AGI | Not measured | 90.2% | Only Grok 4.7 measured |
| ARC-AGI-2 | Not measured | 61.4% | Only Grok 4.7 measured |
| Chess puzzles | Not measured | 41.0% | Only Grok 4.7 measured |
| Mystery game puzzles | Not measured | 29.0% | Only Grok 4.7 measured |
| LMArena Text: hard prompts | Not measured | 1415 | Only Grok 4.7 measured |
| LMArena Text: expert | Not measured | 1439 | Only Grok 4.7 measured |
| LMArena Text: hard prompts in English | Not measured | 1427 | Only Grok 4.7 measured |
| LMArena Vision: diagrams | Not measured | 1254 | Only Grok 4.7 measured |
| LMArena Vision: homework | Not measured | 1309 | Only Grok 4.7 measured |
| Neither model has a result on the other 8 boards here | |||
MathDirectional onlyPublic boards, shown as published, kept out of the index6 boards measured
| Benchmark | Grok 4.20 Multi-Agent | Grok 4.7 | Leader |
|---|---|---|---|
| OTIS Mock AIME | Not measured | 98.1% | Only Grok 4.7 measured |
| FrontierMath, tiers 1 to 3 | Not measured | 53.0% | Only Grok 4.7 measured |
| FrontierMath, tier 4 | Not measured | 17.1% | Only Grok 4.7 measured |
| ProofBench | Not measured | 26.0% | Only Grok 4.7 measured |
| LMArena Text: math | Not measured | 1414 | Only Grok 4.7 measured |
| LMArena Text: mathematical work | Not measured | 1435 | Only Grok 4.7 measured |
| Neither model has a result on the other 7 boards here | |||
KnowledgeDirectional onlyPublic boards, shown as published, kept out of the index7 boards measured
| Benchmark | Grok 4.20 Multi-Agent | Grok 4.7 | Leader |
|---|---|---|---|
| LegalBench (Vals AI) | Not measured | 84.4% | Only Grok 4.7 measured |
| MedCode (Vals AI) | Not measured | 49.5% | Only Grok 4.7 measured |
| SimpleQA Verified | Not measured | 56.0% | Only Grok 4.7 measured |
| LMArena Text: science | Not measured | 1419 | Only Grok 4.7 measured |
| LMArena Text: medicine and healthcare | Not measured | 1395 | Only Grok 4.7 measured |
| LMArena Text: legal and government | Not measured | 1423 | Only Grok 4.7 measured |
| LMArena Text: business and finance | Not measured | 1389 | Only Grok 4.7 measured |
| Neither model has a result on the other 10 boards here | |||
Web designDirectional onlyPublic boards, shown as published, kept out of the index7 boards measured
| Benchmark | Grok 4.20 Multi-Agent | Grok 4.7 | Leader |
|---|---|---|---|
| Design Arena: websites | Not measured | 1254 | Only Grok 4.7 measured |
| Design Arena: UI components | Not measured | 1275 | Only Grok 4.7 measured |
| Design Arena: game dev | Not measured | 1313 | Only Grok 4.7 measured |
| Design Arena: data viz | Not measured | 1235 | Only Grok 4.7 measured |
| Design Arena: 3D | Not measured | 1270 | Only Grok 4.7 measured |
| Design Arena: all code categories | Not measured | 1271 | Only Grok 4.7 measured |
| LiveBench design | Not measured | 78.0% | Only Grok 4.7 measured |
| Neither model has a result on the other 2 boards here | |||
GeneralDirectional onlyPublic boards, shown as published, kept out of the index31 boards measured
| Benchmark | Grok 4.20 Multi-Agent | Grok 4.7 | Leader |
|---|---|---|---|
| Vals Index | Not measured | 58.6% | Only Grok 4.7 measured |
| LiveBench data analysis | Not measured | 76.9% | Only Grok 4.7 measured |
| Public Benefits Bench (Vals AI) | Not measured | 65.6% | Only Grok 4.7 measured |
| SAGE (Vals AI) | Not measured | 40.8% | Only Grok 4.7 measured |
| MedScribe (Vals AI) | Not measured | 89.4% | Only Grok 4.7 measured |
| Long context reasoning (AA-LCR) | Not measured | 80.0% | Only Grok 4.7 measured |
| DTBench | Not measured | 96.0% | Only Grok 4.7 measured |
| LMCA | Not measured | 49.4% | Only Grok 4.7 measured |
| Furniture assembly | Not measured | 20.8% | Only Grok 4.7 measured |
| LMArena Text | Not measured | 1399 | Only Grok 4.7 measured |
| LMArena Text: instruction following | Not measured | 1406 | Only Grok 4.7 measured |
| LMArena Text: creative writing | Not measured | 1381 | Only Grok 4.7 measured |
| LMArena Vision | Not measured | 1231 | Only Grok 4.7 measured |
| LMArena Text: English | Not measured | 1411 | Only Grok 4.7 measured |
| LMArena Text: non-English | Not measured | 1386 | Only Grok 4.7 measured |
| LMArena Text: Chinese | Not measured | 1449 | Only Grok 4.7 measured |
| LMArena Text: German | Not measured | 1366 | Only Grok 4.7 measured |
| LMArena Text: Japanese | Not measured | 1359 | Only Grok 4.7 measured |
| LMArena Text: Korean | Not measured | 1383 | Only Grok 4.7 measured |
| LMArena Text: French | Not measured | 1431 | Only Grok 4.7 measured |
| LMArena Text: Spanish | Not measured | 1406 | Only Grok 4.7 measured |
| LMArena Text: Russian | Not measured | 1399 | Only Grok 4.7 measured |
| LMArena Text: Polish | Not measured | 1420 | Only Grok 4.7 measured |
| LMArena Text: multi-turn | Not measured | 1394 | Only Grok 4.7 measured |
| LMArena Text: longer queries | Not measured | 1416 | Only Grok 4.7 measured |
| LMArena Text: writing and language | Not measured | 1394 | Only Grok 4.7 measured |
| LMArena Text: entertainment and media | Not measured | 1376 | Only Grok 4.7 measured |
| LMArena Vision: English | Not measured | 1229 | Only Grok 4.7 measured |
| LMArena Vision: Chinese | Not measured | 1250 | Only Grok 4.7 measured |
| LMArena Vision: reading text | Not measured | 1241 | Only Grok 4.7 measured |
| LMArena Vision: creative writing | Not measured | 1217 | Only Grok 4.7 measured |
| Neither model has a result on the other 15 boards here | |||
What each workload costs
Three fixed token mixes at each model's listed price, so per-token rates become a bill.
Chat turn
1K input + 500 output tokens
Grok 4.20 Multi-Agent costs less
Repository review
50K input + 3K output tokens
Grok 4.20 Multi-Agent costs less
Agent loop
200K input + 10K output tokens
Grok 4.20 Multi-Agent costs less
Listed OpenRouter rates, without cache or batch discounts. A workload that does not fit a model's context window would need splitting.
Specification differences
The leading value in a row is marked; a row with a missing value names no leader. Our two rows are marked in red.
| Specification | SpaceXAI Grok 4.20 Multi-Agent | SpaceXAI Grok 4.7 |
|---|---|---|
| Maker | xAI | xAI |
| Weights | No public weights listed | No public weights listed |
| Released | March 31, 2026 | September 21, 2026 |
| Context window | 2M | 500K |
| Input $/Mtok | $1.25 | $2.00 |
| Output $/Mtok | $2.50 | $6.00 |
| Coding index | Not benchmarked | 66.7 |
| Independent measurements behind it | 0 | 5 |
| Cost per solved task | Not computable | Not computable |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is Grok 4.20 Multi-Agent better than Grok 4.7 for coding?
Only Grok 4.7 has a coding score, 66.7 on the Vibe Coding Index; Grok 4.20 Multi-Agent has none, so the two cannot be ranked on coding. Pick Grok 4.20 Multi-Agent for output tokens 2.4x cheaper ($2.50 vs $6.00 per million) and a larger context window (2M against 500K tokens).
Which is cheaper, Grok 4.20 Multi-Agent or Grok 4.7?
Grok 4.20 Multi-Agent costs $1.25 per million input tokens and $2.50 per million output tokens; Grok 4.7 costs $2.00 and $6.00. Cost per solved task is known for neither of them.
Which has the bigger context window, Grok 4.20 Multi-Agent or Grok 4.7?
Grok 4.20 Multi-Agent, with 2M tokens against 500K.
Has any benchmark tested both Grok 4.20 Multi-Agent and Grok 4.7?
No benchmark has tested both yet. Grok 4.20 Multi-Agent has none yet; Grok 4.7 on LiveBench coding, LMArena WebDev and Terminal-Bench 4 (Vals AI). Compare them on price and context instead. Data as of October 21, 2026.