LLMs TigerLive LLM benchmarks

Qwen3.8-27B (XHigh): benchmark scores, price and ranking

Qwen3.8-27B (XHigh) from Alibaba ranks #348 of 743 LLMs on the LLMs Tiger aggregate benchmark score (48.5), based on 4 benchmark results captured up to 2026-10-13.

Organisation
Alibaba
Type
Open-weight / local model
Open weights
Yes
Released
2026-08-14 (source: epoch)
Cheapest API price
$0.040 in / $2.30 out per million tokens (via openrouter, observed 2026-10-13T20:47Z)
Aggregate rank
#348 of 743 (score 48.5)

Relative to other models, Qwen3.8-27B (XHigh) is strongest on CritPt (#48 of 115) and weakest on Surface Evolver (#8 of 13).

Across the 4 benchmarks where at least ten models are scored, Qwen3.8-27B (XHigh) is in the top 10% on 0 and in the top half on 3.

Its cheapest listed input price, $0.040 per million tokens, is lower than that of 94% of the 427 models with a listed price.

Qwen3.8-27B (XHigh) benchmark results

Qwen3.8-27B (XHigh) benchmark scores, ranked against all models measured on each benchmark
BenchmarkScoreRankSourceCaptured
CritPt5.4%#48 of 115Epoch AI Benchmarking Hub2026-10-13
ARC-AGI-242.4%#39 of 87ARC Prize official leaderboard2026-10-13
SciCode46.6%#49 of 106Epoch AI Benchmarking Hub2026-10-13
Surface Evolver45.0%#8 of 13Epoch AI Benchmarking Hub2026-10-13

About Qwen3.8-27B (XHigh)

Qwen3.8-27B (XHigh) is an open-weight model that can run on your own hardware from Alibaba. It was released on 2026-08-14, according to Epoch AI. The cheapest API listing we track charges $0.040 per million input tokens and $2.30 per million output tokens, through openrouter. We hold 4 benchmark results for it across 3 areas, and none is in the top tenth of its board.

Where Qwen3.8-27B (XHigh) stands

On our aggregate score, Qwen3.8-27B (XHigh) is #348 of 743 ranked models, around the middle of the table. Its score is 48.5; the scale is centred on 50, so it sits 1.5 points below the typical model. That beats 53% of the ranked field. Among the 74 ranked Alibaba models it is #35. The next one up is qwen2.5-coder:7b-instruct-q8_0 at 48.5. The next one down is qwen3.5-35b-a3b at 48.2.

Qwen3.8-27B (XHigh), benchmark by benchmark

CritPt puts Qwen3.8-27B (XHigh) at 5.4%, #48 of 115. That beats 58% of the models tested, but 26.9 points short of GPT-5.6 Sol (max) (32.3%). The results either side are Inkling (xhigh) (5.4%) and GLM-5.1 (4.6%). It covers unpublished research-level physics problems. The board median is 2.9%, so Qwen3.8-27B (XHigh) is 2.6 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.

On ARC-AGI-2, Qwen3.8-27B (XHigh) reaches 42.4%. It ranks #39 of 87, ahead of 55% of the field, and 52.6 points behind GPT-6 Astra (max) at 95.0%. The results either side are Gemini 3 Deep Think (Preview) ² (45.1%) and Inkling Small (xhigh) (40.1%). It covers abstract reasoning, ARC Prize public eval set. The board median is 22.8%, so Qwen3.8-27B (XHigh) is 19.6 points above it. Source: ARC Prize official leaderboard, captured 2026-10-13.

Qwen3.8-27B (XHigh) manages 46.6% on SciCode, good for #49 of 106. Claude Opus 5.5 (max) leads at 66.9%, 20.3 points further up, and 54% of models sit below. The results either side are Claude Sonnet 4.6 (max) (46.8%) and Qwen3.7 Plus (45.5%). It covers scientific research code. The board median is 44.3%, so Qwen3.8-27B (XHigh) is 2.3 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.

Qwen3.8-27B (XHigh) lands below the middle on Surface Evolver: 45.0%, #8 of 13, with 54% of the field scoring higher. Kimi K3 (max) leads at 93.0%; the closest result above is DeepSeek V4.1 Flash (high) at 46.2%. It covers geometry optimisation with a solver. The board median is 46.2%, so Qwen3.8-27B (XHigh) is 1.2 points below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.

Strengths and weaknesses

Reasoning, maths and science: solid, ahead of 57% of the field on average across 2 benchmarks. Its best board here is CritPt (#48 of 115) and its weakest is ARC-AGI-2 (#39 of 87).

Price and value

The median input price across the 427 priced models is $0.500 per million tokens, so Qwen3.8-27B (XHigh) costs less 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), DeepSeek V4 Flash 0731 (max) ($0.016, 53.1). Similar scores cost more on Llama-3.2-11B-Vision-Instruct ($0.049) and granite-4.1-8b ($0.050).

Release history

The Alibaba model released before it was Qwen3.8 Max (xhigh) on 2026-08-02, which scores 53.2. The next one was Qwen3.8 Flash (xhigh) on 2026-08-26, at 48.9.

Reading these numbers

The 4 scores on this page come from 2 sources: Epoch AI Benchmarking Hub (3), ARC Prize official leaderboard (1). 1 of them come from an official leaderboard and 3 from datasets or third-party runs. The leaderboard also lists 5 other reasoning-effort settings of this model, which redirect here. This page shows the highest. With 4 results the aggregate rank can move a lot when one new score arrives.

Qwen3.8-27B (XHigh) is also listed at 5 other reasoning-effort settings; this page shows the highest-effort row, which is the one the leaderboard keeps by default.

Models ranked nearby

More from Alibaba

Compare it on the live leaderboard → · How the aggregate score works