Multi-Dimensional Public Ranked Directories (233 Verified Entries)
Inspect MCP servers, Cloud Ollama & frontier agent models, benchmarks, engineering/security/humanitarian RSS feeds, and verified harness files via HTTPS (GET /v1/public/lists/:slug) or MCP (teleon_public_list_query).
top-agent-models (40) top-benchmarks (36) top-harness-files (45) top-mcp-servers (52) top-rss-feeds (60)
| Rank | Entry & Summary | Category | Composite | Measured Dimensions |
|---|---|---|---|---|
| #1 | DeepSeek V3.2 (Cloud Ollama) Served via https://ollama.com/v1 with MIT weights, strong multi-tool reliability, and high SWE task efficiency. |
ollama_cloud_v1 |
88.64 | {"coding_swe_verified_pct":69.4,"context_32k_fidelity":94.5,"cost_efficiency_score":95.8,"jagged_frontier_stability":88.9,"recommended_profile":"high_value_open_weights_swe_worker","tool_call_reliability":94.6} |
| #2 | Claude Sonnet 4.6 Workhorse repository engineering and tool-use model with strong perturbed-prompt stability. |
frontier_cloud_api |
88.34 | {"coding_swe_verified_pct":74.8,"context_32k_fidelity":96.8,"cost_efficiency_score":81.5,"jagged_frontier_stability":91.2,"recommended_profile":"primary_multi_file_swe_agent","tool_call_reliability":97.4} |
| #3 | Qwen3-Coder 30B (qwen3-coder:30b on Cloud Ollama) MoE coding specialist on https://ollama.com/v1 delivering near-frontier repository edits at fraction-cent cost. |
ollama_cloud_v1 |
87.96 | {"coding_swe_verified_pct":66.8,"context_32k_fidelity":93.6,"cost_efficiency_score":97.4,"jagged_frontier_stability":87.8,"recommended_profile":"fast_opencode_leaf_builder","tool_call_reliability":94.2} |
| #4 | Moonshot Kimi K2 (Cloud Ollama) 1T-parameter MoE trained specifically for multi-turn MCP tool invocation and autonomous terminal loops via https://ollama.com/v1. |
ollama_cloud_v1 |
87.58 | {"coding_swe_verified_pct":67.5,"context_32k_fidelity":94.1,"cost_efficiency_score":92.6,"jagged_frontier_stability":88.4,"recommended_profile":"long_horizon_mcp_shell_agent","tool_call_reliability":95.3} |
| #5 | Gemma 4 31B (gemma4:31b on Cloud Ollama) Demonstrates +0.062 Pareto non-regression lift under <=2048B pre-turn Contract Cards on https://ollama.com/v1. |
ollama_cloud_v1 |
87.22 | {"coding_swe_verified_pct":62.4,"context_32k_fidelity":93.8,"cost_efficiency_score":94.9,"jagged_frontier_stability":91.6,"recommended_profile":"pareto_contract_card_executor","tool_call_reliability":93.4} |
| #6 | OpenAI GPT-5 High-reliability multimodal and multi-step tool-calling frontier baseline. |
frontier_cloud_api |
87.18 | {"coding_swe_verified_pct":74.9,"context_32k_fidelity":97.1,"cost_efficiency_score":76.2,"jagged_frontier_stability":90.8,"recommended_profile":"complex_orchestration_and_synthesis","tool_call_reliability":96.9} |
| #7 | OpenAI o4-mini Fast reasoning model for test generation, SQL tie-outs, and tool-assisted quantitative checks. |
frontier_cloud_api |
87.18 | {"coding_swe_verified_pct":68.1,"context_32k_fidelity":95,"cost_efficiency_score":89.4,"jagged_frontier_stability":88.6,"recommended_profile":"cost_bounded_unit_test_verifier","tool_call_reliability":94.8} |
| #8 | Gemini 2.5 Pro Excels at whole-repository context digestion, cross-module invariant audits, and multimodal trace inspection. |
frontier_cloud_api |
87.12 | {"coding_swe_verified_pct":70.6,"context_32k_fidelity":98.2,"cost_efficiency_score":82,"jagged_frontier_stability":89.7,"recommended_profile":"long_context_codegraph_auditor","tool_call_reliability":95.1} |
| #9 | Qwen3 235B-A22B (Cloud Ollama) 22B-active sparse MoE on https://ollama.com/v1 combining strong multilingual reasoning with Apache-2.0 portability. |
ollama_cloud_v1 |
87.1 | {"coding_swe_verified_pct":65.9,"context_32k_fidelity":94.7,"cost_efficiency_score":91.8,"jagged_frontier_stability":89.2,"recommended_profile":"open_weights_blind_consensus_judge","tool_call_reliability":93.9} |
| #10 | Gemini 2.5 Flash Controllable thinking budget and fast TTFT ideal for parallel feed qualification and document extraction. |
frontier_cloud_api |
87.04 | {"coding_swe_verified_pct":63.5,"context_32k_fidelity":96,"cost_efficiency_score":94.2,"jagged_frontier_stability":87.9,"recommended_profile":"high_throughput_subagent_worker","tool_call_reliability":93.6} |
| #11 | Gemma 4 27B (gemma4:27b on Cloud Ollama) High-stability dense model on https://ollama.com/v1 for schema validation, SPDX audits, and low-variance scoring. |
ollama_cloud_v1 |
86.64 | {"coding_swe_verified_pct":60.8,"context_32k_fidelity":93.1,"cost_efficiency_score":95.6,"jagged_frontier_stability":90.9,"recommended_profile":"structured_extraction_and_rubric_gate","tool_call_reliability":92.8} |
| #12 | Zhipu GLM-4.5 (Cloud Ollama) Unified reasoning, coding, and function-calling open model served through https://ollama.com/v1. |
ollama_cloud_v1 |
86.26 | {"coding_swe_verified_pct":64.2,"context_32k_fidelity":92.9,"cost_efficiency_score":93.8,"jagged_frontier_stability":87.2,"recommended_profile":"hybrid_reasoning_tool_agent","tool_call_reliability":93.2} |
| #13 | OpenAI GPT-5 mini Low-cost structured output and parallel tool-calling model for high-volume queue consumers. |
frontier_cloud_api |
86.16 | {"coding_swe_verified_pct":61.5,"context_32k_fidelity":94.2,"cost_efficiency_score":93.9,"jagged_frontier_stability":87.4,"recommended_profile":"parallel_queue_batch_worker","tool_call_reliability":93.8} |
| #14 | Mistral Devstral (Cloud Ollama) Co-designed for open SWE scaffolds via https://ollama.com/v1 with compact tool call tokens and Apache-2.0 licensing. |
ollama_cloud_v1 |
86.12 | {"coding_swe_verified_pct":61.2,"context_32k_fidelity":92.4,"cost_efficiency_score":96.4,"jagged_frontier_stability":87.5,"recommended_profile":"swe_scaffold_patch_runner","tool_call_reliability":93.1} |
| #15 | Claude Haiku 4.5 Fast subagent exploration, codebase search triage, and guardrail verification. |
frontier_cloud_api |
85.68 | {"coding_swe_verified_pct":59.8,"context_32k_fidelity":93.9,"cost_efficiency_score":93.2,"jagged_frontier_stability":88.1,"recommended_profile":"low_latency_subagent_scout","tool_call_reliability":93.4} |
| #16 | Qwen3 32B Dense (qwen3:32b on Cloud Ollama) Dense 32B model on https://ollama.com/v1 offering predictable latency and faithful SQLite/Postgres DDL synthesis. |
ollama_cloud_v1 |
85.54 | {"coding_swe_verified_pct":59.4,"context_32k_fidelity":92.8,"cost_efficiency_score":94.7,"jagged_frontier_stability":88.3,"recommended_profile":"dense_verifier_and_sql_generator","tool_call_reliability":92.5} |
| #17 | Claude Opus 4.6 Top-tier architectural refactoring, formal specification review, and blind conflict arbitration. |
frontier_cloud_api |
85.48 | {"coding_swe_verified_pct":76.2,"context_32k_fidelity":97.6,"cost_efficiency_score":62.4,"jagged_frontier_stability":93.4,"recommended_profile":"deep_architecture_and_3judge_arbiter","tool_call_reliability":97.8} |
| #18 | Mistral Medium 3 (Cloud Ollama) Balanced enterprise coding and strict JSON tool-use model accessible via https://ollama.com/v1 and Mistral APIs. |
ollama_cloud_v1 |
85 | {"coding_swe_verified_pct":60.1,"context_32k_fidelity":93,"cost_efficiency_score":91.2,"jagged_frontier_stability":88,"recommended_profile":"enterprise_api_connector_agent","tool_call_reliability":92.7} |
| #19 | Muse Spark Open (muse-spark-open on Cloud Ollama) Fast open synthesis and editorial brief generator running on https://ollama.com/v1 with zero em-dash discipline. |
ollama_cloud_v1 |
84.9 | {"coding_swe_verified_pct":57.6,"context_32k_fidelity":91.8,"cost_efficiency_score":96.1,"jagged_frontier_stability":87.6,"recommended_profile":"editorial_synthesis_and_blueprint_cards","tool_call_reliability":91.4} |
| #20 | OpenAI o3 Search-heavy reasoning model suited for subtle concurrency bugs, Lean 4 proof sketches, and algorithmic edge cases. |
frontier_cloud_api |
84.86 | {"coding_swe_verified_pct":71.7,"context_32k_fidelity":96.4,"cost_efficiency_score":68.5,"jagged_frontier_stability":92.1,"recommended_profile":"formal_verification_and_root_cause_search","tool_call_reliability":95.6} |
| #21 | xAI Grok Code Fast 1 High token-per-second coding loop specialist tuned for rapid grep-edit-test cycles. |
frontier_cloud_api |
84.82 | {"coding_swe_verified_pct":60.4,"context_32k_fidelity":91.6,"cost_efficiency_score":94.5,"jagged_frontier_stability":85.7,"recommended_profile":"high_speed_iterative_test_loop","tool_call_reliability":91.9} |
| #22 | Meta Llama 4 Maverick (Cloud Ollama) 17B-active 128-expert multimodal MoE on https://ollama.com/v1 for fast UI screenshot + JSON routing. |
ollama_cloud_v1 |
84.64 | {"coding_swe_verified_pct":58.4,"context_32k_fidelity":92.6,"cost_efficiency_score":93.5,"jagged_frontier_stability":86.9,"recommended_profile":"multimodal_triage_and_routing","tool_call_reliability":91.8} |
| #23 | Qwen 2.5 Coder 32B (qwen2.5-coder:32b on Cloud Ollama) Mature Apache-2.0 coding baseline on https://ollama.com/v1 with highly reproducible TAEDRI primitive output. |
ollama_cloud_v1 |
84.2 | {"coding_swe_verified_pct":55.2,"context_32k_fidelity":91.5,"cost_efficiency_score":95.4,"jagged_frontier_stability":87.9,"recommended_profile":"deterministic_sql_and_python_primitive_gen","tool_call_reliability":91} |
| #24 | DeepSeek R1 Distill 70B (deepseek-r1:70b on Cloud Ollama) Open reasoning checkpoint on https://ollama.com/v1 strong at adversarial falsification and math/logic checks. |
ollama_cloud_v1 |
84.04 | {"coding_swe_verified_pct":58.9,"context_32k_fidelity":92.2,"cost_efficiency_score":91.5,"jagged_frontier_stability":89,"recommended_profile":"invariant_counterexample_finder","tool_call_reliability":88.6} |
| #25 | NVIDIA Llama-3.3 Nemotron Super 49B (Cloud Ollama) NAS-optimized 49B model on https://ollama.com/v1 with toggleable reasoning traces for throughput control. |
ollama_cloud_v1 |
83.98 | {"coding_swe_verified_pct":56.8,"context_32k_fidelity":92,"cost_efficiency_score":92.8,"jagged_frontier_stability":87.1,"recommended_profile":"steerable_reasoning_judge","tool_call_reliability":91.2} |
| #26 | MiniMax-M1 456B Hybrid lightning-attention open-weights reasoning model handling 80k-token CI/CD failure logs cleanly. |
open_weights_cloud |
83.72 | {"coding_swe_verified_pct":56,"context_32k_fidelity":95.4,"cost_efficiency_score":90.6,"jagged_frontier_stability":86.2,"recommended_profile":"long_trace_failure_postmortem","tool_call_reliability":90.4} |
| #27 | OpenAI GPT-4.1 Non-reasoning 1M-context model with strong literal instruction adherence and exact patch formatting. |
frontier_cloud_api |
83.66 | {"coding_swe_verified_pct":54.6,"context_32k_fidelity":96.5,"cost_efficiency_score":84.6,"jagged_frontier_stability":88.5,"recommended_profile":"long_context_literal_diff_applier","tool_call_reliability":94.1} |
| #28 | ByteDance Seed-OSS-36B-Instruct Apache-2.0 36B agent model supporting explicit token-budget caps on chain-of-thought steps. |
open_weights_cloud |
83.64 | {"coding_swe_verified_pct":55.6,"context_32k_fidelity":92.9,"cost_efficiency_score":92.4,"jagged_frontier_stability":86.5,"recommended_profile":"budget_capped_thinking_worker","tool_call_reliability":90.8} |
| #29 | Cohere Command A 111B (command-a:111b on Cloud Ollama) Built for verifiable inline source citations and multi-tool enterprise RAG workflows via https://ollama.com/v1. |
ollama_cloud_v1 |
83.62 | {"coding_swe_verified_pct":54.1,"context_32k_fidelity":94.1,"cost_efficiency_score":88.9,"jagged_frontier_stability":88.4,"recommended_profile":"citation_grounded_rag_synthesis","tool_call_reliability":92.6} |
| #30 | xAI Grok 4 Provides cross-family model diversity in 3-judge blind reconciliation panels. |
frontier_cloud_api |
83.48 | {"coding_swe_verified_pct":66.2,"context_32k_fidelity":94,"cost_efficiency_score":78.4,"jagged_frontier_stability":86.8,"recommended_profile":"independent_blind_judge_diversity","tool_call_reliability":92} |
| #31 | Meta Llama 4 Scout (Cloud Ollama) 16-expert long-context scout model on https://ollama.com/v1 for multi-file source qualification. |
ollama_cloud_v1 |
83.34 | {"coding_swe_verified_pct":52.4,"context_32k_fidelity":93.4,"cost_efficiency_score":95.2,"jagged_frontier_stability":85.9,"recommended_profile":"long_document_foraging_scout","tool_call_reliability":89.8} |
| #32 | GLM-5.3-Flash (glm-5.3-flash on Cloud Ollama) Ultra-fast Cloud Ollama (https://ollama.com/v1) classifier for L0 scent cards; Pareto-gated against unassisted complex diffs. |
ollama_cloud_v1 |
83.1 | {"coding_swe_verified_pct":53.8,"context_32k_fidelity":90.2,"cost_efficiency_score":98.6,"jagged_frontier_stability":82.4,"recommended_profile":"rung1_fast_scent_classifier_only","tool_call_reliability":90.5} |
| #33 | Mistral Codestral 22B (codestral:22b on Cloud Ollama) Low-latency code completion and unit-test scaffold generator served via https://ollama.com/v1. |
ollama_cloud_v1 |
83.02 | {"coding_swe_verified_pct":51.9,"context_32k_fidelity":90.8,"cost_efficiency_score":95.9,"jagged_frontier_stability":86.8,"recommended_profile":"low_latency_fim_and_unit_tests","tool_call_reliability":89.7} |
| #34 | Gemini 2.5 Flash-Lite Ultra-low-cost high-RPS classifier for RSS feed deduplication, taxonomy tagging, and TTL triage. |
frontier_cloud_api |
82.34 | {"coding_swe_verified_pct":46.2,"context_32k_fidelity":92.1,"cost_efficiency_score":99.1,"jagged_frontier_stability":84.8,"recommended_profile":"high_volume_rss_classification","tool_call_reliability":89.5} |
| #35 | Mistral Large 2411 Strong multilingual instruction following and parallel function calling across European languages. |
frontier_cloud_api |
82.06 | {"coding_swe_verified_pct":54,"context_32k_fidelity":92.7,"cost_efficiency_score":84.2,"jagged_frontier_stability":87.3,"recommended_profile":"multilingual_legal_spdx_reviewer","tool_call_reliability":92.1} |
| #36 | Microsoft Phi-4 Reasoning 14B (phi-4-reasoning:14b on Cloud Ollama) Compact MIT-licensed 14B reasoning model on https://ollama.com/v1 for constraint and arithmetic checks. |
ollama_cloud_v1 |
81.94 | {"coding_swe_verified_pct":48.5,"context_32k_fidelity":89.4,"cost_efficiency_score":97.8,"jagged_frontier_stability":86.1,"recommended_profile":"compact_math_and_constraint_checker","tool_call_reliability":87.9} |
| #37 | Amazon Nova Pro v1 Multimodal document and tool workflow runner with IAM-scoped Bedrock deployment controls. |
frontier_cloud_api |
80.44 | {"coding_swe_verified_pct":45.8,"context_32k_fidelity":91.7,"cost_efficiency_score":90.4,"jagged_frontier_stability":85.1,"recommended_profile":"aws_vpc_document_workflow_runner","tool_call_reliability":89.2} |
| #38 | AllenAI OLMo 2 32B Instruct (olmo-2:32b on Cloud Ollama) Fully open weights, training data, and recipes via https://ollama.com/v1 enabling benchmark contamination audits. |
ollama_cloud_v1 |
80.16 | {"coding_swe_verified_pct":43.5,"context_32k_fidelity":89.8,"cost_efficiency_score":94,"jagged_frontier_stability":86.7,"recommended_profile":"contamination_free_scientific_baseline","tool_call_reliability":86.8} |
| #39 | IBM Granite 3.3 8B Instruct (granite-3.3:8b on Cloud Ollama) Compact Apache-2.0 enterprise governance classifier on https://ollama.com/v1 with indemnified training provenance. |
ollama_cloud_v1 |
80.08 | {"coding_swe_verified_pct":41.2,"context_32k_fidelity":88.9,"cost_efficiency_score":98.4,"jagged_frontier_stability":85.5,"recommended_profile":"apache2_governance_and_pii_gate","tool_call_reliability":86.4} |
| #40 | Perplexity Sonar Reasoning Pro Live web-grounded verification specialist for checking upstream pricing changes and deprecation notices. |
search_grounded_api |
78.54 | {"coding_swe_verified_pct":44,"context_32k_fidelity":91.2,"cost_efficiency_score":83.5,"jagged_frontier_stability":85.6,"recommended_profile":"live_fragility_and_deprecation_verifier","tool_call_reliability":88.4} |