Benchmark Almanac

LiveNewsBench Release 2

DeepSeek-V3.2 84.5% leads since January 2026
Not in model cards
Version Release 2

Models search for answers to questions about recent news. The test limits searches and page visits and compares answers with curated references.

Braintrust independently ran LiveNewsBench on 1,329 questions. The authors publish distinct test files for the September and January releases.

Released
January 2026
Kind
Questions that need search
Level
Everyday
Scored
Graded by a model
Answers
Short answer
Tasks
All public
Measures
Capability
Language
English
In and out
Text → Text
Interaction
Agentic
Environment
Browser
Items from
Real work and use
Run by
Anyone
Based on
LiveNewsBench
In views
Agents
Strongest model tested
GPT-5.2 #51 on the capability index
Last model added
January 2026
Scores refreshed
November 21st, 2026
13
Models scored on its own leaderboard

DeepSeek-V3.2 leads with 84.5% among models released by January 2026

Best score per model from the benchmark's own leaderboard, read on November 21st, 2026. Where a model ran with several settings or harnesses, we keep its best. Every run is in the next chart.

The newest model with a score here came out in January 2026, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

  1. 1 DeepSeek-V3.2 Standard Settings (5 Searches, 5 Visits); reasoning yes; open yes; official search no 84.5%
  2. 2 Grok 4 Standard Settings (5 Searches, 5 Visits); reasoning yes; open no; official search no 82%
  3. 3 Claude Sonnet 4.5 Standard Settings (5 Searches, 5 Visits); reasoning no; open no; official search no 82%
  4. 4 Gemini 3 Pro 7 Searches, 5 Visits; reasoning yes; open no; official search no 74.5%
  5. 5 GPT-5.2 7 Searches, 5 Visits; reasoning yes; open no; official search no 74.5%
  6. 6 GPT-4.1 Standard Settings (5 Searches, 5 Visits); reasoning no; open no; official search no 72.5%
  7. 7 Kimi K2.5 Standard Settings (5 Searches, 5 Visits); reasoning yes; open yes; official search no 70.5%
  8. 8 Qwen3-235B-A22B-Instruct Jul 2025, Standard Settings (5 Searches, 5 Visits); reasoning no; open yes; official search no 61.5%
  9. 9 Qwen3-235B-A22B-Thinking Jul 2025, Standard Settings (5 Searches, 5 Visits); reasoning yes; open yes; official search no 60%
  10. 10 Kimi K2 Thinking 7 Searches, 5 Visits; reasoning yes; open yes; official search no 52%
  11. 11 Qwen3-8B Standard Settings (5 Searches, 5 Visits); reasoning no; open yes; official search no 49.5%
  12. 12 gpt-oss-120b 7 Searches, 5 Visits; reasoning yes; open yes; official search no 29.5%
  13. 13 Llama 3.1-8B Standard Settings (5 Searches, 5 Visits); reasoning no; open yes; official search no 11%

Source: Official LiveNewsBench Release 2 leaderboard

#ModelScoreSettingDateSource
1 DeepSeek-V3.2 84.5% Standard Settings (5 Searches, 5 Visits); reasoning yes; open yes; official search no Dec 2025 Official leaderboard
2 Grok 4 82% Standard Settings (5 Searches, 5 Visits); reasoning yes; open no; official search no Jul 2025 Official leaderboard
3 Claude Sonnet 4.5 82% Standard Settings (5 Searches, 5 Visits); reasoning no; open no; official search no Sep 2025 Official leaderboard
4 Gemini 3 Pro 74.5% 7 Searches, 5 Visits; reasoning yes; open no; official search no Nov 2025 Official leaderboard
5 GPT-5.2 74.5% 7 Searches, 5 Visits; reasoning yes; open no; official search no Dec 2025 Official leaderboard
6 GPT-4.1 72.5% Standard Settings (5 Searches, 5 Visits); reasoning no; open no; official search no Apr 2025 Official leaderboard
7 Kimi K2.5 70.5% Standard Settings (5 Searches, 5 Visits); reasoning yes; open yes; official search no Jan 2026 Official leaderboard
8 Qwen3-235B-A22B-Instruct Jul 2025 61.5% Standard Settings (5 Searches, 5 Visits); reasoning no; open yes; official search no Jul 2025 Official leaderboard
9 Qwen3-235B-A22B-Thinking Jul 2025 60% Standard Settings (5 Searches, 5 Visits); reasoning yes; open yes; official search no Jul 2025 Official leaderboard
10 Kimi K2 Thinking 52% 7 Searches, 5 Visits; reasoning yes; open yes; official search no Nov 2025 Official leaderboard
11 Qwen3-8B 49.5% Standard Settings (5 Searches, 5 Visits); reasoning no; open yes; official search no Apr 2025 Official leaderboard
12 gpt-oss-120b 29.5% 7 Searches, 5 Visits; reasoning yes; open yes; official search no Aug 2025 Official leaderboard
13 Llama 3.1-8B 11% Standard Settings (5 Searches, 5 Visits); reasoning no; open yes; official search no Jul 2024 Official leaderboard

Scores at each effort level

Models the leaderboard ran at two or more effort levels. We don’t have enough cost data here to say what the extra effort costs.

Lowest effort to highest, light to dark
DeepSeek-V3.2
1 Search, 5… 1 Search, 5… 3 Searches,…
84.5%
Claude Sonnet 4.5
1 Search, 5… 7 Searches,… 3 Searches,…
82%
Gemini 3 Pro
1 Search, 5… Standard Se… 3 Searches,…
74.5%
GPT-5.2
1 Search, 5… 3 Searches,…
74.5%
GPT-4.1
1 Search, 5… Standard Se…
72.5%
Kimi K2 Thinking
1 Search, 5… 3 Searches,…
52%
Qwen3-8B
1 Search, 5… 3 Searches,…
49.5%
gpt-oss-120b
1 Search, 5… Standard Se…
29.5%

Source: Official LiveNewsBench Release 2 leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
DeepSeek-V3.2 Standard Settings (5 Searches, 5 Visits); reasoning yes; open yes; official search no 84.5%
DeepSeek-V3.2 7 Searches, 5 Visits; reasoning no; open yes; official search no 84.5%
DeepSeek-V3.2 7 Searches, 5 Visits; reasoning yes; open yes; official search no 84.5%
DeepSeek-V3.2 Standard Settings (5 Searches, 5 Visits); reasoning no; open yes; official search no 83%
Grok 4 Standard Settings (5 Searches, 5 Visits); reasoning yes; open no; official search no 82%
Claude Sonnet 4.5 Standard Settings (5 Searches, 5 Visits); reasoning no; open no; official search no 82%
DeepSeek-V3.2 3 Searches, 5 Visits; reasoning yes; open yes; official search no 80.5%
Claude Sonnet 4.5 3 Searches, 5 Visits; reasoning no; open no; official search no 79%
DeepSeek-V3.2 3 Searches, 5 Visits; reasoning no; open yes; official search no 79%
Gemini 3 Pro 7 Searches, 5 Visits; reasoning yes; open no; official search no 74.5%
GPT-5.2 7 Searches, 5 Visits; reasoning yes; open no; official search no 74.5%
GPT-5.2 Standard Settings (5 Searches, 5 Visits); reasoning yes; open no; official search no 74%
GPT-4.1 Standard Settings (5 Searches, 5 Visits); reasoning no; open no; official search no 72.5%
GPT-5.2 3 Searches, 5 Visits; reasoning yes; open no; official search no 72.5%
Kimi K2.5 Standard Settings (5 Searches, 5 Visits); reasoning yes; open yes; official search no 70.5%
Gemini 3 Pro 3 Searches, 5 Visits; reasoning yes; open no; official search no 70%
GPT-4.1 3 Searches, 5 Visits; reasoning no; open no; official search no 69%
Claude Sonnet 4.5 7 Searches, 5 Visits; reasoning no; open no; official search no 67%
GPT-4.1 7 Searches, 5 Visits; reasoning no; open no; official search no 66.5%
GPT-4.1 1 Search, 5 Visits; reasoning no; open no; official search no 62.5%
GPT-5.2 1 Search, 5 Visits; reasoning yes; open no; official search no 62.5%
Qwen3-235B-A22B-Instruct Jul 2025 Standard Settings (5 Searches, 5 Visits); reasoning no; open yes; official search no 61.5%
Gemini 3 Pro Standard Settings (5 Searches, 5 Visits); reasoning yes; open no; official search no 60.5%
Qwen3-235B-A22B-Thinking Jul 2025 Standard Settings (5 Searches, 5 Visits); reasoning yes; open yes; official search no 60%
Claude Sonnet 4.5 1 Search, 5 Visits; reasoning no; open no; official search no 53.5%
Kimi K2 Thinking 7 Searches, 5 Visits; reasoning yes; open yes; official search no 52%
Qwen3-8B Standard Settings (5 Searches, 5 Visits); reasoning no; open yes; official search no 49.5%
Qwen3-8B 7 Searches, 5 Visits; reasoning no; open yes; official search no 49.5%
DeepSeek-V3.2 1 Search, 5 Visits; reasoning yes; open yes; official search no 48.5%
Qwen3-8B 3 Searches, 5 Visits; reasoning no; open yes; official search no 48%
Kimi K2 Thinking Standard Settings (5 Searches, 5 Visits); reasoning yes; open yes; official search no 48%
Kimi K2 Thinking 3 Searches, 5 Visits; reasoning yes; open yes; official search no 47%
gpt-oss-120b 7 Searches, 5 Visits; reasoning yes; open yes; official search no 29.5%
Gemini 3 Pro 1 Search, 5 Visits; reasoning yes; open no; official search no 29.5%
gpt-oss-120b 3 Searches, 5 Visits; reasoning yes; open yes; official search no 28.5%
gpt-oss-120b Standard Settings (5 Searches, 5 Visits); reasoning yes; open yes; official search no 27%
Qwen3-8B 1 Search, 5 Visits; reasoning no; open yes; official search no 22%
DeepSeek-V3.2 1 Search, 5 Visits; reasoning no; open yes; official search no 20%
Llama 3.1-8B Standard Settings (5 Searches, 5 Visits); reasoning no; open yes; official search no 11%
Kimi K2 Thinking 1 Search, 5 Visits; reasoning yes; open yes; official search no 7.50%
gpt-oss-120b 1 Search, 5 Visits; reasoning yes; open yes; official search no 6.00%

Number one over time

Lab claims count from the day the card came out, independent runs from the model’s release.

DeepSeek
Own leaderboard Not tested since Jan 2026

Source: Official LiveNewsBench Release 2 leaderboard

Scores over time

All the scores we have, one dot each, so a model tested at several settings appears more than once.

Plot with other benchmarks
On the benchmark’s own leaderboard Best so far
0% 20% 40% 60% 80% 100% 2025 2026 Near the ceiling (85%) DeepSeek-V3.2 Grok 4 GPT-4.1 Llama 3.1-8B 0% 20% 40% 60% 80% 100% 2025 2026 Near the ceiling (85%) DeepSeek-V3.2 GPT-4.1

Source: Official LiveNewsBench Release 2 leaderboard

No lab model card reports it yet

We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.

Versions

We switch the top row to each new version when it comes out. Open it for one row per version. An old version stays in play while people still test it, so whoever first takes it past 85% gets credit for closing it out.

OpenAI Google DeepMind DeepSeek
LiveNewsBench Release 2 Not tested since Jan 2026
LiveNewsBench Replaced by Release 2, Jan 2026
GPT-5.2 21.5%
LiveNewsBench Release 2 Not tested since Jan 2026

Source: Official leaderboards

More in Knowledge

All 332 in Knowledge
  • MMLU-Pro In use Harder 10-option version of MMLU with reasoning-focused questions across 14 subjects. Intern-S2-397B 89.8%
  • MMLU In use 57-subject multiple-choice exam questions from elementary to professional level. Scored by accuracy. Seed1.8 92.3%
  • SuperGPQA In use 26,529 graduate-level multiple-choice questions across 285 disciplines. Accuracy. Qwen 3.6 Max (Preview) 73.9%
  • WideSearch In use 200 bilingual broad info-seeking tasks: agent collects many atomic facts into a table. Scored by success rate and row/item F1. Tencent Hy4 preview 83.9%
  • BBEH micro average In use Models solve harder replacements for each of the 23 BIG-Bench Hard tasks, 4,520 examples in total. The score is the share of all examples answered correctly. Gemma 4 31B IT 74.4%
  • HLE-Diamond In use Models answer 1,000 cleaned expert questions from Humanity's Last Exam, 500 on reasoning and 500 on knowledge, closed book. Graded on exact answers. Claude Fable 5.2 77.9%

Data audit

9 of 10 checks pass, 1 warning. Look first at the Newest score check. The newest came 10 months ago, in January 2026, for Kimi K2.5. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 1 warning 3 pass 1 note
  • Pass Independent runs Scores from the official leaderboard (13 entries).
  • Pass Models scored 13 models we track.
  • Warning Newest score The newest came 10 months ago, in January 2026, for Kimi K2.5.
  • Note Frontier models None of the top 10 has a score. The strongest tested is GPT-5.2 (#51).
  • Pass Headroom The best is 84.5%, under 85%, so it still separates models.
Leaderboard import 3 pass
  • Pass Importer We read the board on livenewsbench.com: 41 rows for models we track at the last read.
  • Pass Last read November 21st, 2026.
  • Pass Read worked The last read worked.
Consistency 1 pass 1 note
  • Note Lab claims No model has both a lab number and an independent run.
  • Pass Known issues No published audit found problems.

How we check