grok-4-20: benchmark scores, price and ranking
grok-4-20 from xAI ranks #218 of 743 LLMs on the LLMs Tiger aggregate benchmark score (50.0), based on 11 benchmark results captured up to 2026-10-13.
- Organisation
- xAI
- Type
- Frontier model
- Open weights
- No
- Released
- 2026-02-17 (source: epoch)
- Cheapest API price
- $1.25 in / $2.50 out per million tokens (via openrouter, observed 2026-10-13T19:47Z)
- Aggregate rank
- #218 of 743 (score 50.0)
Relative to other models, grok-4-20 is strongest on Epoch ECI (#6 of 204) and weakest on WebDev Arena (#42 of 67).
Across the 9 benchmarks where at least ten models are scored, grok-4-20 is in the top 10% on 1 and in the top half on 8.
Its cheapest listed input price, $1.25 per million tokens, is lower than that of 26% of the 427 models with a listed price.
grok-4-20 benchmark results
| Benchmark | Score | Rank | Source | Captured |
|---|---|---|---|---|
| Epoch ECI | 152 | #6 of 204 | Epoch AI Benchmarking Hub | 2026-10-13 |
| CL-bench | 22.2% | #3 of 16 | Epoch AI Benchmarking Hub | 2026-10-13 |
| DTBench | 90.1% | #28 of 146 | Epoch AI Benchmarking Hub | 2026-10-13 |
| Terminal-Bench 2.0 | 57.3% | #7 of 28 | Epoch AI Benchmarking Hub | 2026-10-13 |
| WeirdML | 52.3% | #29 of 109 | Epoch AI Benchmarking Hub | 2026-10-13 |
| ALE-Bench | 1,150 | #22 of 79 | Epoch AI Benchmarking Hub | 2026-10-13 |
| LMCA | 38.7% | #32 of 108 | Epoch AI Benchmarking Hub | 2026-10-13 |
| Vending-Bench 2 | $4,663 | #7 of 19 | Epoch AI Benchmarking Hub | 2026-10-13 |
| CL-bench Life | 11.9% | #5 of 9 | Epoch AI Benchmarking Hub | 2026-10-13 |
| WebDev Arena | 1,374 | #42 of 67 | Epoch AI Benchmarking Hub | 2026-10-13 |
| Blueprint-Bench 2 | 0.0% | #3 of 3 | Epoch AI Benchmarking Hub | 2026-10-13 |
About grok-4-20
grok-4-20 is a frontier model from xAI. It was released on 2026-02-17, according to Epoch AI. The cheapest API listing we track charges $1.25 per million input tokens and $2.50 per million output tokens, through openrouter. We hold 11 benchmark results for it across 3 areas, and 1 is in the top tenth of its board.
Where grok-4-20 stands
On our aggregate score, grok-4-20 is #218 of 743 ranked models, around the middle of the table. Its score is 50.0; the scale is centred on 50, so it sits 0.0 points above the typical model. That beats 71% of the ranked field. Among the 26 ranked xAI models it is #10. The next one up is grok-3-beta at 50.0. The next one down is grok-4.20-0309-reasoning at 49.6.
grok-4-20, benchmark by benchmark
On Epoch ECI, grok-4-20 scores 152 and ranks #6 of 204. Only 5 models score higher, led by Qwen 3.8 Max at 156, which is 4 points ahead. It covers Epoch Capabilities Index: Epoch's own aggregate of ~40 benchmarks. The board median is 127, so grok-4-20 is 25 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
On CL-bench, grok-4-20 reaches 22.2%. It ranks #3 of 16, ahead of 81% of the field, and 5.7 points behind GPT-5.4 (xhigh) at 27.9%. The results either side are GPT-5.1 (high) (23.7%) and Qwen 3.6 Plus (2026-04-02) (20.3%). It covers learning new rules from context. The board median is 17.7%, so grok-4-20 is 4.5 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
grok-4-20 manages 90.1% on DTBench, good for #28 of 146. Claude Opus 5.5 (max) leads at 98.9%, 8.8 points further up, and 81% of models sit below. The results either side are GPT-6 Luna (max) (90.1%) and nemotron-3-ultra (90.1%). It covers dynamic tool-use benchmark. The board median is 68.8%, so grok-4-20 is 21.3 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
Terminal-Bench 2.0 puts grok-4-20 at 57.3%, #7 of 28. That beats 75% of the models tested, but 12.1 points short of Gemini 3 Pro Preview (69.4%). The results either side are GPT-5.1 Codex (57.8%) and GLM-5 (52.4%). It covers real terminal tasks (agent + model). The board median is 37.3%, so grok-4-20 is 20.0 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
On WeirdML, grok-4-20 reaches 52.3%. It ranks #29 of 109, ahead of 73% of the field, and 41.0 points behind GPT-6 Astra (max) at 93.3%. The results either side are Gemma 4 31B IT (52.3%) and DeepSeek v4 Pro (max) (48.9%). It covers unusual machine-learning tasks. The board median is 41.2%, so grok-4-20 is 11.1 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
grok-4-20 manages 1,150 on ALE-Bench, good for #22 of 79. GPT-6 Astra (max) leads at 2,951, 1,801 points further up, and 72% of models sit below. The results either side are mimo-v2.6-pro (1,158) and Kimi K2.6 (1,093). It covers AtCoder heuristic-optimisation contests (performance rating). The board median is 764, so grok-4-20 is 386 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
LMCA puts grok-4-20 at 38.7%, #32 of 108. That beats 70% of the models tested, but 29.6 points short of Claude Opus 5.5 (max) (68.2%). The results either side are Gemma 4 31B IT (39.2%) and grok-4.3 (high) (38.3%). It covers language-model capability assessment. The board median is 29.2%, so grok-4-20 is 9.4 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
On Vending-Bench 2, grok-4-20 reaches $4,663. It ranks #7 of 19, ahead of 63% of the field, and 1,542 points behind Kimi K2.6 at $6,205. The results either side are Claude Fable 5 (max) ($4,967) and GLM-5 ($4,432). It covers running a vending business for a year (final money, $). The board median is $2,377, so grok-4-20 is 2,286 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
grok-4-20 lands below the middle on WebDev Arena: 1,374, #42 of 67, with 61% of the field scoring higher. Claude Opus 5.5 (max) leads at 1,820; the closest result above is MiniMax-M2.5 at 1,384. It covers building web apps, crowd-voted (Elo). The board median is 1,433, so grok-4-20 is 60 points below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
The full table above also lists the other 2 results, including boards with fewer than ten models, where a rank says little.
Strengths and weaknesses
Software engineering: solid, ahead of 61% of the field on average across 3 benchmarks. Its best board here is Terminal-Bench 2.0 (#7 of 28) and its weakest is WebDev Arena (#42 of 67). General knowledge and chat: strong, ahead of 77% of the field on average across 5 benchmarks. Its best board here is Epoch ECI (#6 of 204) and its weakest is Vending-Bench 2 (#7 of 19).
Price and value
The median input price across the 427 priced models is $0.500 per million tokens, so grok-4-20 costs more than the typical listing. Models that score within 1.5 points or better and cost less: Qwen2.5-Coder-7B-Instruct ($0.010, 49.5), DeepSeek V4.1 Flash (high) ($0.010, 48.9), Qwen3.7 Flash ($0.030, 50.5). Similar scores cost more on glm-5p1 ($1.40) and Qwen3.8 Max (0902) (xhigh) ($1.65).
Release history
The xAI model released before it was grok-4-1 on 2025-11-17, which scores 47.5. The next one was grok-4.20-0309-reasoning on 2026-02-17, at 49.6.
Reading these numbers
The 11 scores on this page come from 1 source: Epoch AI Benchmarking Hub (11). No official leaderboard supplied any of them, so treat them as third-party measurements.
Models ranked nearby
- #215 glm-4.7-flash (50.1)
- #216 grok-3-beta (50.0)
- #217 Claude Sonnet 4.6 (max) (50.0)
- #219 step-5-preview-high (50.0)
- #220 claude-sonnet-4-20250514_2K (50.0)
- #221 gemini-exp-1114 (50.0)
More from xAI
- Grok 4.6 (xhigh) (#34, 53.6)
- grok-4-0709 (#38, 53.2)
- Grok 4 (#39, 53.2)
- Grok 4.7 (xhigh) (#41, 53.1)
- Grok 4.5 (high) (#85, 52.0)
- Grok 4.20 (Reasoning) (#136, 51.2)
Compare it on the live leaderboard → · How the aggregate score works