Resyst Labs Benchmarks

Independent model evaluation · public evidence · tactical replays

AI model benchmarks with rules, receipts, and Arena replays.

Resyst Labs Benchmarks is a public evidence surface for comparing AI systems across agentic reasoning, software engineering, reliability, runtime economics, and Resyst Arena duels.

Explore the ranking Enter Resyst Arena
Current leader GPT‑5.5 Overall 88.22 · OpenRouter full/SWE · OpenAI-compatible Hard Intelligence
20 ranked entrants
5 Arena matches
Jul 09, 2026 data refresh
01

Agentic discipline

Structured outputs, tool-use boundaries, instruction following, grounded reasoning, and hallucination resistance.

02

Software execution

Practical implementation quality, final-answer usefulness, source handling, and architecture cleanliness.

03

Hard Intelligence

Active inquiry, online adaptation, self-repair, and authority integrity under a public hard-reasoning diagnostic lane.

04

Resyst Arena

Turn-based spatial duels where legal action discipline and tactical continuity are measured separately from runtime telemetry.

Unified ranking

One table, visible tradeoffs.

Local and API-backed systems share a single tournament view. Provider, runtime, cost, latency, reliability, and lane basis remain visible so comparisons stay honest.

Open ranking explainer Download ranking JSON
#1

GPT‑5.5

OpenRouter full/SWE · OpenAI-compatible Hard Intelligence

Open result → 88.2
#2

GPT‑5.6 Terra

OpenRouter · extra-high reasoning · Full + SWE + Hard measured

Open result → 87.7
#3

DeepSeek V4 Flash

DeepSeek direct API · refreshed Hard Intelligence telemetry

Open result → 86.1
Rank Model Basis Overall Full SWE Hard IQ Inquiry Adapt Repair Authority Cost Reliability Details
#1 GPT‑5.5 OpenRouter full/SWE · OpenAI-compatible Hard Intelligence 88.22 85.76 85.51 93.40 93.08 98.13 82.40 100.00 $4.071 100.0% Result
#2 GPT‑5.6 Terra OpenRouter · extra-high reasoning · Full + SWE + Hard measured 87.70 88.85 86.08 88.18 99.75 97.50 85.48 70.00 $1.867 100.0% Result
#3 DeepSeek V4 Flash DeepSeek direct API · refreshed Hard Intelligence telemetry 86.08 95.80 86.95 75.48 80.74 81.70 73.28 66.21 $0.290 100.0% Result
#4 GPT‑5.6 Sol OpenRouter · extra-high reasoning · Full + SWE + Hard measured 84.57 86.27 76.12 91.31 99.75 80.00 85.48 100.00 $3.858 100.0% Result
#5 Claude Opus 4.8 OpenRouter · extra-high reasoning · Hard Intelligence 84.45 83.32 88.67 81.37 93.08 80.00 82.40 70.00 $6.115 100.0% Result
#6 Claude Fable 5 OpenRouter · extra-high reasoning · Full + SWE + Hard measured 83.74 77.02 81.42 92.78 99.75 98.13 85.73 87.50 $12.285 100.0% Result
#7 GLM‑5.2 OpenRouter · z-ai/glm-5.2 · maximum reasoning · Full + SWE + Hard measured 83.68 86.87 89.58 74.58 17.50 98.13 82.71 100.00 $2.768 100.0% Result
#8 Gemini 3.5 Flash OpenRouter · extra-high reasoning · Hard Intelligence 83.38 88.79 73.59 87.75 81.42 97.50 82.08 90.00 $1.646 100.0% Result
#9 DeepSeek V4 Pro DeepSeek direct API · maximum-reasoning Hard IQ 83.36 92.04 79.68 78.38 81.42 80.00 82.08 70.00 $0.335 100.0% Result
#10 GPT‑5.6 Luna OpenRouter · extra-high reasoning · Full + SWE + Hard measured 82.13 89.90 75.84 80.66 99.75 80.00 85.40 57.50 $1.000 100.0% Result
#11 Claude Sonnet 5 OpenRouter · extra-high reasoning · Full + SWE + Hard measured 82.08 81.64 78.08 86.53 98.50 99.06 86.04 62.50 $2.812 100.0% Result
#12 Qwen3.7 Max OpenRouter · extra-high reasoning · Hard Intelligence 80.71 83.24 88.99 69.90 17.50 80.00 82.08 100.00 $0.906 100.0% Result
#13 MiniMax M3 OpenRouter · extra-high reasoning 79.66 85.33 86.88 66.77 17.50 97.50 82.08 70.00 $0.182 100.0% Result
#14 MiniMax M3 Direct Plus MiniMax Plus · direct API · SWE extra-high reasoning 74.63 84.32 67.77 71.80 17.50 98.50 71.19 100.00 $0.0041 100.0% Result
#15 Kimi K2.7 Code OpenRouter · Kimi K2.7 Code · extra-high Hard IQ 71.01 87.95 58.61 66.46 17.50 80.00 78.33 90.00 $0.732 100.0% Result
#16 Step 3.7 Flash OpenRouter · stepfun/step-3.7-flash · extra-high reasoning · Full + SWE + Hard measured 68.06 89.79 80.39 33.99 17.50 80.00 29.46 9.00 $0.494 100.0% Result
#17 NVIDIA Nemotron 3 Ultra OpenRouter · nvidia/nemotron-3-ultra-550b-a55b · extra-high reasoning 67.58 79.54 58.63 64.56 93.08 8.75 56.40 100.00 $0.489 100.0% Result
#18 Gemma4‑12B‑Coder Fable5/Composer2.5 Q4_K_M Local model Local GGUF · llama.cpp Vulkan · Q4_K_M 61.19 80.60 54.01 48.96 17.50 6.25 82.08 90.00 $0 100.0% Result
#19 Ornith‑1.0‑35B Q4_K_M Local model Local GGUF · llama.cpp Vulkan · Q4_K_M · 35B MoE 55.57 77.96 39.16 49.58 18.75 11.25 78.33 90.00 $0 100.0% Result
#20 Qwythos‑9B Claude Mythos Q8_0 Local model Local GGUF · llama.cpp Vulkan · Q8_0 · 256K allocation verified 55.13 74.98 46.51 43.91 81.42 6.88 78.33 9.00 $0 100.0% Result
Resyst Arena

A tactical testbed, not a latency race.

Resyst Arena evaluates spatial strategy in deterministic turn-based games. Each public match summary links to a replay with board states, legal actions, events, and tactical telemetry.

  • Side-swapped series before ranking-grade claims.
  • Legal action rate and invalid actions are first-class evidence.
  • Replay data stays linked to tactical summaries.
Open the Arena replay room
Methodology

Scores are claims with receipts.

Separate lanes

Agentic, software-engineering, and Hard Intelligence diagnostics are preserved as distinct measurements before any publication formula combines them.

Runtime honesty

The same model can appear through different providers or runtimes. The table exposes basis metadata instead of hiding infrastructure differences.

Evidence thresholds

Single runs are evidence records. Stronger claims require repeated series, side swaps, seed variation, and comparable scoring settings.

Evidence contract

Public by design. Auditable by default.

This site publishes benchmark summaries as versioned data files. The presentation layer is intentionally separate from the scoring harness so rankings can evolve without rewriting the public record.

Ranking data Arena data