Benchmark Almanac

Swiss Smartvote 2023 party agreement

Claude 3.7 Sonnet 80.1% leads since February 2025
Not in model cards

Models answer 75 Swiss election questions about their own policy positions. Researchers compare their answers with political parties and candidates to measure agreement.

Lucerne researchers ran three models 200 times each. Barmettler compared 66 models with 184 elected National Council members. Agreement measures proximity to each political reference.

Made by
Smartvote and Jannis Vamvas
Released
August 2023
Kind
Bias, fairness and politics
Level
Everyday
Scored
Checked automatically
Answers
Multiple choice
Tasks
All public
Measures
Propensity
Behavior
Bias
Language
German
In and out
Text → Text
Interaction
Single turn
Items from
Real work and use
Refresh
One fixed set
Run by
Anyone
Strongest model tested
GPT-5.4 #28 on the capability index
Last model added
March 2026
Scores refreshed
November 20th, 2026
65
Entries on its own leaderboard, 51 models we track

Claude 3.7 Sonnet leads with 80.1% among models released by March 2026 Claude 3.7 Sonnet leads with 80.1% among models released by March 2026

Best score per entry from the benchmark's own leaderboard, read on November 20th, 2026. The source gives no overall score. This is mean agreement with six parties, with equal weight for each party. Where a model ran with several settings or harnesses, we keep its best. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders. This view shows only models we track, each at its best.

The newest model with a score here came out in March 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 Claude 3.7 Sonnet 80.1%
  2. 2 Amazon Nova Premier 80%
  3. 3 Claude Opus 4 80%
  4. 4 Claude Opus 4.6 79.9%
  5. 5 Mistral Medium 3 79.7%
  6. 6 Gemini 2.0 Flash Feb 2025 79.7%
  7. 7 Yi-Large 79.7%
  8. 8 GLM-5 79.7%
  9. 9 Claude 3.5 Sonnet (snapshot unspecified) 79.6%
  10. 10 Phi-4 79.5%
  11. 11 MiMo-V2-Pro 79.5%
  12. 12 Grok 4.20 79.5%
  13. 13 Mistral Small 4 79.5%
  14. 14 GPT-3.5 Turbo (snapshot unspecified) 79.3%
  15. 15 Mistral Small 3 (Jan 2025) 79.3%
  16. 16 Mixtral 8x22B 79.2%
  17. 17 Claude 3 Haiku 79%
  18. 18 GPT-5 79%
  19. 19 Jamba 1.5 Large 79%
  20. 20 GPT-4 Turbo (snapshot unspecified) 78.9%
  21. 21 Mistral Large (Dec 2025) 78.8%
  22. 22 Nova Pro 78.8%
  23. 23 GPT-5.4 Mini 78.7%
  24. 24 Mistral NeMo 78.7%
  25. 25 Kimi K2.5 78.7%
  26. 26 Qwen3.5-Plus 78.6%
  27. 27 Doubao-Seed-1.6 78.6%
  28. 28 o3-mini 78.6%
  29. 29 Qwen3.5 397B-A17B 78.6%
  30. 30 Llama 4 Maverick 78.6%
  31. 31 Claude Sonnet 4 78.5%
  32. 32 Grok 3 78.5%
  33. 33 Command A 78.5%
  34. 34 DeepSeek-V3.2 78.4%
  35. 35 DeepSeek v3 (snapshot unspecified) 78.4%
  36. 36 GPT-5.4 78.4%
  37. 37 Qwen2.5-72B 78.4%
  38. 38 Grok-2 Dec 2024 78.3%
  39. 39 o4-mini 78.2%
  40. 40 Claude Sonnet 4.5 78.2%
  41. 41 GPT-4o (snapshot unspecified) 78.2%
  42. 42 Command R+ 78.2%
  43. 43 Mistral Large (Nov 2024) 78%
  44. 44 Claude Sonnet 4.6 78%
  45. 45 Hermes 3 Llama 3.1 405B 77.9%
  46. 46 Claude 3.5 Haiku 77.7%
  47. 47 GPT-4.1 77.6%
  48. 48 Hermes 4 405B 77.6%
  49. 49 Qwen3-Max 77.6%
  50. 50 DeepSeek R1 (snapshot unspecified) 77.5%
  51. 51 o1 77.5%
  52. 52 o1-mini 77.4%
  53. 53 Grok 4 Fast 77.2%
  54. 54 Minimax 77.2%
  55. 55 Grok-3 mini 77%
  56. 56 DeepSeek-R1 May 2025 76.8%
  57. 57 MiniMax-M2.7 76.7%
  58. 58 GPT-4o Mini (snapshot unspecified) 76.6%
  59. 59 Mistral Large 76.3%
  60. 60 Grok 4 75.8%
  61. 61 Grok 4.1 Fast 74.3%
  62. 62 Llama 3.1-405B 74.3%
  63. 63 Qwen 3.6 Plus 71.5%
  64. 64 Mistral-7B-Instruct-v0.1 69.4%
  65. 65 Nemotron 3 Super 61.8%

Source: Official Swiss Smartvote 2023 party agreement leaderboard

Source: Official Swiss Smartvote 2023 party agreement leaderboard

#EntryScoreSettingDateSource
1 Claude 3.7 Sonnet 80.1% Feb 2025 Official leaderboard
2 Amazon Nova Premier 80% Apr 2025 Official leaderboard
3 Claude Opus 4 80% May 2025 Official leaderboard
4 Claude Opus 4.6 79.9% Feb 2026 Official leaderboard
5 Mistral Medium 3 79.7% May 2025 Official leaderboard
6 Gemini 2.0 Flash Feb 2025 79.7% Feb 2025 Official leaderboard
7 Yi-Large 79.7% May 2024 Official leaderboard
8 GLM-5 79.7% Feb 2026 Official leaderboard
9 Claude 3.5 Sonnet (snapshot unspecified) 79.6% Official leaderboard
10 Phi-4 79.5% Dec 2024 Official leaderboard
11 MiMo-V2-Pro 79.5% Mar 2026 Official leaderboard
12 Grok 4.20 79.5% Feb 2026 Official leaderboard
13 Mistral Small 4 79.5% Mar 2026 Official leaderboard
14 GPT-3.5 Turbo (snapshot unspecified) 79.3% Official leaderboard
15 Mistral Small 3 (Jan 2025) 79.3% Official leaderboard
16 Mixtral 8x22B 79.2% Apr 2024 Official leaderboard
17 Claude 3 Haiku 79% Mar 2024 Official leaderboard
18 GPT-5 79% Aug 2025 Official leaderboard
19 Jamba 1.5 Large 79% Aug 2024 Official leaderboard
20 GPT-4 Turbo (snapshot unspecified) 78.9% Official leaderboard
21 Mistral Large (Dec 2025) 78.8% Official leaderboard
22 Nova Pro 78.8% Official leaderboard
23 GPT-5.4 Mini 78.7% Mar 2026 Official leaderboard
24 Mistral NeMo 78.7% Jul 2024 Official leaderboard
25 Kimi K2.5 78.7% Jan 2026 Official leaderboard
26 Qwen3.5-Plus 78.6% Feb 2026 Official leaderboard
27 Doubao-Seed-1.6 78.6% Jun 2025 Official leaderboard
28 o3-mini 78.6% Jan 2025 Official leaderboard
29 Qwen3.5 397B-A17B 78.6% Feb 2026 Official leaderboard
30 Llama 4 Maverick 78.6% Apr 2025 Official leaderboard
31 Claude Sonnet 4 78.5% May 2025 Official leaderboard
32 Grok 3 78.5% Apr 2025 Official leaderboard
33 Command A 78.5% Official leaderboard
34 DeepSeek-V3.2 78.4% Dec 2025 Official leaderboard
35 DeepSeek v3 (snapshot unspecified) 78.4% Official leaderboard
36 GPT-5.4 78.4% Mar 2026 Official leaderboard
37 Qwen2.5-72B 78.4% Sep 2024 Official leaderboard
38 Grok-2 Dec 2024 78.3% Dec 2024 Official leaderboard
39 o4-mini 78.2% Apr 2025 Official leaderboard
40 Claude Sonnet 4.5 78.2% Sep 2025 Official leaderboard
41 GPT-4o (snapshot unspecified) 78.2% Official leaderboard
42 Command R+ 78.2% Aug 2024 Official leaderboard
43 Mistral Large (Nov 2024) 78% Official leaderboard
44 Claude Sonnet 4.6 78% Feb 2026 Official leaderboard
45 Hermes 3 Llama 3.1 405B 77.9% Official leaderboard
46 Claude 3.5 Haiku 77.7% Oct 2024 Official leaderboard
47 GPT-4.1 77.6% Apr 2025 Official leaderboard
48 Hermes 4 405B 77.6% Aug 2025 Official leaderboard
49 Qwen3-Max 77.6% Sep 2025 Official leaderboard
50 DeepSeek R1 (snapshot unspecified) 77.5% Official leaderboard
51 o1 77.5% Dec 2024 Official leaderboard
52 o1-mini 77.4% Sep 2024 Official leaderboard
53 Grok 4 Fast 77.2% Sep 2025 Official leaderboard
54 Minimax 77.2% Official leaderboard
55 Grok-3 mini 77% Apr 2025 Official leaderboard
56 DeepSeek-R1 May 2025 76.8% May 2025 Official leaderboard
57 MiniMax-M2.7 76.7% Mar 2026 Official leaderboard
58 GPT-4o Mini (snapshot unspecified) 76.6% Official leaderboard
59 Mistral Large 76.3% Feb 2024 Official leaderboard
60 Grok 4 75.8% Jul 2025 Official leaderboard
61 Grok 4.1 Fast 74.3% Nov 2025 Official leaderboard
62 Llama 3.1-405B 74.3% Jul 2024 Official leaderboard
63 Qwen 3.6 Plus 71.5% Mar 2026 Official leaderboard
64 Mistral-7B-Instruct-v0.1 69.4% Sep 2023 Official leaderboard
65 Nemotron 3 Super 61.8% Mar 2026 Official leaderboard
#ModelScoreSettingDateSource
1 Claude 3.7 Sonnet 80.1% Feb 2025 Official leaderboard
2 Amazon Nova Premier 80% Apr 2025 Official leaderboard
3 Claude Opus 4 80% May 2025 Official leaderboard
4 Claude Opus 4.6 79.9% Feb 2026 Official leaderboard
5 Mistral Medium 3 79.7% May 2025 Official leaderboard
6 Gemini 2.0 Flash Feb 2025 79.7% Feb 2025 Official leaderboard
7 Yi-Large 79.7% May 2024 Official leaderboard
8 GLM-5 79.7% Feb 2026 Official leaderboard
9 Phi-4 79.5% Dec 2024 Official leaderboard
10 MiMo-V2-Pro 79.5% Mar 2026 Official leaderboard
11 Grok 4.20 79.5% Feb 2026 Official leaderboard
12 Mistral Small 4 79.5% Mar 2026 Official leaderboard
13 Mixtral 8x22B 79.2% Apr 2024 Official leaderboard
14 Claude 3 Haiku 79% Mar 2024 Official leaderboard
15 GPT-5 79% Aug 2025 Official leaderboard
16 Jamba 1.5 Large 79% Aug 2024 Official leaderboard
17 GPT-5.4 Mini 78.7% Mar 2026 Official leaderboard
18 Mistral NeMo 78.7% Jul 2024 Official leaderboard
19 Kimi K2.5 78.7% Jan 2026 Official leaderboard
20 Qwen3.5-Plus 78.6% Feb 2026 Official leaderboard
21 Doubao-Seed-1.6 78.6% Jun 2025 Official leaderboard
22 o3-mini 78.6% Jan 2025 Official leaderboard
23 Qwen3.5 397B-A17B 78.6% Feb 2026 Official leaderboard
24 Llama 4 Maverick 78.6% Apr 2025 Official leaderboard
25 Claude Sonnet 4 78.5% May 2025 Official leaderboard
26 Grok 3 78.5% Apr 2025 Official leaderboard
27 DeepSeek-V3.2 78.4% Dec 2025 Official leaderboard
28 GPT-5.4 78.4% Mar 2026 Official leaderboard
29 Qwen2.5-72B 78.4% Sep 2024 Official leaderboard
30 Grok-2 Dec 2024 78.3% Dec 2024 Official leaderboard
31 o4-mini 78.2% Apr 2025 Official leaderboard
32 Claude Sonnet 4.5 78.2% Sep 2025 Official leaderboard
33 Command R+ 78.2% Aug 2024 Official leaderboard
34 Claude Sonnet 4.6 78% Feb 2026 Official leaderboard
35 Claude 3.5 Haiku 77.7% Oct 2024 Official leaderboard
36 GPT-4.1 77.6% Apr 2025 Official leaderboard
37 Hermes 4 405B 77.6% Aug 2025 Official leaderboard
38 Qwen3-Max 77.6% Sep 2025 Official leaderboard
39 o1 77.5% Dec 2024 Official leaderboard
40 o1-mini 77.4% Sep 2024 Official leaderboard
41 Grok 4 Fast 77.2% Sep 2025 Official leaderboard
42 Grok-3 mini 77% Apr 2025 Official leaderboard
43 DeepSeek-R1 May 2025 76.8% May 2025 Official leaderboard
44 MiniMax-M2.7 76.7% Mar 2026 Official leaderboard
45 Mistral Large 76.3% Feb 2024 Official leaderboard
46 Grok 4 75.8% Jul 2025 Official leaderboard
47 Grok 4.1 Fast 74.3% Nov 2025 Official leaderboard
48 Llama 3.1-405B 74.3% Jul 2024 Official leaderboard
49 Qwen 3.6 Plus 71.5% Mar 2026 Official leaderboard
50 Mistral-7B-Instruct-v0.1 69.4% Sep 2023 Official leaderboard
51 Nemotron 3 Super 61.8% Mar 2026 Official leaderboard

Rankings by part

We read 6 parts from the benchmark’s own leaderboard, each ranking models on one part of the test, such as SP, Grüne and GLP. GPT-4.1 leads 2 of them, and 4 other models lead the rest. Arrows show how far a model moved from its overall rank among the models we track.

SP

51 models we track have a score here.

  1. 1 GPT-4.1 ↑35 90.9%
  2. 2 GPT-5.4 ↑26 90.8%
  3. 3 o1 ↑36 90.2%
  4. 4 GPT-5 ↑11 89.3%
  5. 5 Mistral Large ↑40 89.3%
  6. 6 Qwen3-Max ↑32 89.2%
  7. 7 Grok 3 ↑19 89.1%
  8. 8 o4-mini ↑23 89.1%
  9. 9 Amazon Nova Premier ↓7 89%
  10. 10 Claude Opus 4.6 ↓6 89%

7 models have held number one on the official leaderboard since its release in August 2023

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

Mistral AI Anthropic 01.AI Google DeepMind
Own leaderboard Not tested since Mar 2026

Source: Official Swiss Smartvote 2023 party agreement 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% 2024 2025 2026 Near the ceiling (85%) Claude 3.7 Sonnet Yi-Large Mistral-7B-Instruct-v0.1 0% 20% 40% 60% 80% 100% 2024 2025 2026 Near the ceiling (85%) Claude 3.7 Sonnet Mistral-7B-Instruct-v0.1

Source: Official Swiss Smartvote 2023 party agreement 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.

More in Safety and alignment

All 636 in Safety and alignment
  • FORTRESS In use Paired adversarial and benign national-security misuse prompts. Rubric-graded risk score on adversarial and over-refusal score on benign prompts. Claude Opus 5.5 5.78%
  • StrongREJECT In use Forbidden-prompt jailbreak benchmark with a rubric grader that scores how useful a non-refusal actually is. Inkling 98.6%
  • Gray Swan IPI Arena benchmark In use Indirect prompt-injection attacks from public red-teaming competitions replayed on tool, coding and browser agents. Attack success rate at k attempts. Gemini 4 Carbon 0%
  • Covert misbehavior (Thinking Machines) In use Agentic tasks that check whether a model misbehaves and conceals it. Scored by the rate of covert misbehavior. Inkling 2 Mini 8.30%
  • Destructive actions (Thinking Machines) In use Agentic tasks that check whether a model takes destructive actions it was not authorized to take. Scored by the rate of destructive actions. Inkling 2 Small 4.70%
  • Misbehavior (Thinking Machines) In use Agentic tasks that check whether a model fabricates results, ignores instructions or hides what it did when a task is hard. Scored by the rate of misbehavior. Inkling 2 Mini 9.80%

Data audit

9 of 10 checks pass, 1 warning. Look first at the Newest score check. The newest came 8 months ago, in March 2026, for Qwen 3.6 Plus. 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 (65 entries).
  • Pass Models scored 51 models we track, plus 14 other entries on its leaderboard.
  • Warning Newest score The newest came 8 months ago, in March 2026, for Qwen 3.6 Plus.
  • Note Frontier models None of the top 10 has a score. The strongest tested is GPT-5.4 (#28).
  • Pass Headroom The best is 80.1%, under 85%, so it still separates models.
Leaderboard import 3 pass
  • Pass Importer We read the board on github.com: 52 rows for models we track and 14 other entries at the last read.
  • Pass Last read November 20th, 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