{
 "applicability": {
  "basis": "model_type",
  "model_type": "decision-model",
  "not_applicable": [
   "capabilities",
   "inference_performance.api_tps_output",
   "modalities.audio",
   "modalities.document",
   "modalities.embeddings",
   "modalities.image_generation",
   "modalities.reranking",
   "modalities.text.fill_in_middle",
   "modalities.text.json_mode",
   "modalities.text.max_output_tokens",
   "modalities.text.streaming",
   "modalities.video",
   "modalities.vision"
  ]
 },
 "body": "\n# Jev 1.13\n\nJev is TypeSafe's System One model and the first `decision-model` in this\ncatalogue. TypeSafe's documentation describes System One as a class of models\nbuilt to make fast, structured decisions that software can use directly, and\nsays they do not write replies, produce code, or explain their reasoning\n(`docs.typesafe.ai/concepts/system-one.md`, read 2026-09-20). You hand it\nstate, you hand it typed questions, and you get typed answers: a probability\n(**Noul**, 0\u20131), a selected option with its full distribution and a confidence\n(**Choice**, up to 255 options), or an ordered rubric of 2\u201310 levels resolved\nto a probability-weighted value with a confidence (**Score**). Same source,\nsame date.\n\nIt is not an LLM, and the catalogue does not rank it as one. \"Emits no tokens\"\nis the vendor's framing rather than a fact about the wire: the API reports\ninput and output token counts.\n\n## Disclosure\n\n**ModelSpec is a paying TypeSafe customer, and uses this model to judge\nevidence for the catalogue.** `scripts/attribution.py` sends ambiguous creator\nattributions \u2014 \"who actually built the model this page lists?\" \u2014 to Jev, and\npays per input token for the answer (MODEL-82).\n\nSo one rule binds this card and every future TypeSafe card: **no field on a\nTypeSafe card may be written by a Jev judgment.** The refusal is in the code,\nnot in this paragraph. `supplier_conflict()` in `scripts/attribution.py` stops\nthe request before it is made when a supplier is among the candidates, and\n`apply_policy()` refuses a stored judgment that names one, so neither a fresh\ncall nor a cached row can write here. `tests/test_attribution.py` fails if\neither refusal is removed. Deterministic attribution still works \u2014 an id prefix\nthat reads `typesafe/\u2026` is code reading a string, not the supplier's opinion of\nitself \u2014 and anything ambiguous waits for a person.\n\nMoney flows to TypeSafe, not from it. Nothing here is paid placement, a\nreferral fee or provider-paid visibility, and the fact that we buy from a\nvendor buys that vendor nothing in a ranking (see the neutrality commitment).\nThe disclosure is owed anyway, and it is derived from\n`schema/suppliers.py` rather than typed into this file, so the next supplier\ncard cannot quietly omit it.\n\n## Calling it, and what it costs\n\n`POST https://api.typesafe.ai/v1/systemone`, Bearer auth, with a body of\n`{state, model, questions}` (`docs.typesafe.ai/api`, read 2026-09-20). State is\ntext: a string, a JSON object or a JSON array.\n\n**$0.042 per million input tokens. Output tokens are not charged**\n(`docs.typesafe.ai/models.md`, read 2026-09-20). The documented limits, same\nsource and date: 250,000 tokens per **second**, 1,200 requests per minute,\n64,000 tokens per request, of which state plus the longest single question may\nbe at most 32,000.\n\nThe model string is `jev-1.13.0`, with `jev-latest` and `jev-preview` as\naliases; Jev is the only published family member, and TypeSafe documents it as\ntrained with RLCD (`docs.typesafe.ai`, read 2026-09-20). The weights are\nclosed: the official Hugging Face organisation exists and publishes nothing\n(\"None public yet\", `huggingface.co/typesafe`, read 2026-09-20). Reproductions\nby other groups are those groups' artifacts and are not carded as TypeSafe's.\n\n## What this card cannot say\n\nFour researched facts have no field in this schema. They are recorded here, and\nin `cost.note`, rather than smuggled into a field that means something else\n(MODEL-97, MODEL-98):\n\n- **A price per decision.** `cost.input` is per million tokens. A caller of a\n  decision model budgets per judgment \u2014 one request of state plus questions.\n- **A per-second token limit.** `availability.primary_provider.rate_limit_tpm`\n  is per minute. The published limit is 250,000 tokens per second, and\n  multiplying it by 60 would publish a number nobody wrote down, so the field\n  stays null.\n- **The 32k state sub-limit.** `max_input_tokens` holds the 64,000-token\n  request ceiling. The inner limit on state plus the longest question has\n  nowhere to go.\n- **RLCD as a training method.** `lineage.training_method` is an enum without\n  it, so the card says `other` and names RLCD here.\n\nArchitecture is a different kind of blank. Parameters, layers and attention are\n`null` because TypeSafe has not published them, not because the question is\nmeaningless \u2014 that is unknown, and the catalogue should keep asking. What the\nclass genuinely cannot answer \u2014 a maximum output-token count, a token stream,\nan output tok/s \u2014 is derived from `model_type` and shown on the page as \"not\napplicable\", never as a gap (`schema/applicability.py`).\n\n## Evidence, and why it is unranked\n\nTypeSafe publishes **zero benchmark numbers** for Jev. The catalogue therefore\ncarries none, and the ranker returns it `unranked` with\n`insufficient_benchmark_evidence` and a null score. That is the correct\noutcome: a model with no evidence is not a model that scored badly, and the\nrank/unrank split exists precisely so absence is reported rather than priced\nin. The standard suites do not rescue it either \u2014 MMLU, GPQA and SWE-bench\nscore generated answers, and this model generates none.\n\nWhat TypeSafe does publish about failure is a jaggedness page for this version\n(`docs.typesafe.ai/model-jaggedness/jev-1.13.md`, read 2026-09-20): nine\nfailure modes across arithmetic, dates, indirection, adversarial content in the\nsupplied state, and requests to generate rather than judge. No metrics\naccompany them, so `known_failure_modes` carries the areas and nothing is\nscored. Independent evaluations of Jev exist; only their existence was\nverified, not their results, so none is cited.\n\n## Adoption, tested rather than repeated\n\nOne usage figure is verified, and it belongs to one platform: **Vercel's own\nblog (2026-09-18) reports the fastest adoption in its AI Gateway's history \u2014\nroughly 13% of paid AI Gateway teams within 24 hours.** That is a single\nplatform self-reporting about its own customers, which is evidence of real and\nunusually fast uptake on that gateway and is not a market share. No second\nindependent usage source was found: `typesafe-sdk` exists on PyPI, but download\ncounts, repository stars and named customers are not established.\n\nTypeSafe's homepage advertises large speed and cost multiples against\nunnamed comparators. No methodology and no independent measurement accompany\nthem, so this card records no latency, throughput or cost-ratio figure, and\n`inference_performance` is null throughout.\n\n## The organisation\n\nTypeSafe AI, San Francisco, founded 2024; out of stealth on 2026-09-15 \u2014 the\nreports date it 15 or 16 September \u2014 with a $40M seed round led by DCVC. **All\nof that is trade press only** (SiliconANGLE,\nThe Register and heise, read 2026-09-20): `typesafe.ai/blog` returns 404 and no\nfirst-party announcement was found. The exact legal entity is not established\n\u2014 the site footer reads only \"TypeSafe AI \u00a9 2026\" \u2014 so `license_url` and\n`tos_url` are empty rather than guessed, and `commercial_use` is `unspecified`\nrather than inferred from a proprietary API.\n\n## Sources\n\nEvery fact above carries its source and the date it was read. In short:\n\n- `docs.typesafe.ai/concepts/system-one.md` \u2014 what System One is; the three\n  answer types. Read 2026-09-20.\n- `docs.typesafe.ai/api` \u2014 endpoint, auth, request shape. Read 2026-09-20.\n- `docs.typesafe.ai/models.md` \u2014 price, rate and size limits, versions.\n  Read 2026-09-20.\n- `docs.typesafe.ai/model-jaggedness/jev-1.13.md` \u2014 published failure modes,\n  without metrics. Read 2026-09-20.\n- `huggingface.co/typesafe` \u2014 the official organisation, no public weights.\n  Read 2026-09-20.\n- Vercel blog, post dated 2026-09-18 \u2014 the one verified adoption figure,\n  attributed to Vercel. Read 2026-09-20.\n- SiliconANGLE, The Register, heise \u2014 funding, founding and location. Trade\n  press, uncorroborated by any first-party release. Read 2026-09-20.\n",
 "build": {
  "built_at": "2026-09-20T22:59:20+00:00",
  "commit": "6302b8fcd27c077a492f8316eb07f9d894a2ec6a",
  "eligibility_as_of": "2026-09-09",
  "export_schema_version": "3.0"
 },
 "card": {
  "adoption": {
   "community_forks": null,
   "huggingface_downloads": null,
   "huggingface_likes": null,
   "is_common_distillation_teacher": false,
   "is_common_finetune_base": false,
   "notable_users": [],
   "ollama_pulls": null,
   "open_webui_ranking": null,
   "openrouter_ranking": null
  },
  "architecture": {
   "activation_function": "",
   "active_parameters": null,
   "attention_type": null,
   "diffusion_scheduler": "",
   "diffusion_steps_default": null,
   "embedding_dimensions": null,
   "experts_per_token": null,
   "flash_attention": null,
   "hidden_size": null,
   "intermediate_size": null,
   "num_attention_heads": null,
   "num_experts": null,
   "num_kv_heads": null,
   "num_layers": null,
   "positional_encoding": null,
   "precision_native": "",
   "rope_theta": null,
   "sliding_window_size": null,
   "tie_word_embeddings": null,
   "tokenizer_type": null,
   "total_parameters": null,
   "total_parameters_source": "",
   "type": null,
   "vae_type": "",
   "vision_encoder": "",
   "vision_patch_size": null,
   "vision_resolution_max": "",
   "vocab_size": null
  },
  "availability": {
   "ai21_labs": {
    "available": false,
    "fine_tuning": false,
    "gated": false,
    "model_id": "",
    "notes": "",
    "regions": [],
    "url": "https://www.ai21.com/"
   },
   "aws_bedrock": {
    "available": false,
    "fine_tuning": false,
    "gated": false,
    "model_id": "",
    "notes": "",
    "regions": [],
    "url": "https://aws.amazon.com/bedrock/"
   },
   "azure_ai_foundry": {
    "available": false,
    "fine_tuning": false,
    "gated": false,
    "model_id": "",
    "notes": "",
    "regions": [],
    "url": "https://ai.azure.com/"
   },
   "baidu_ernie": {
    "available": false,
    "fine_tuning": false,
    "gated": false,
    "model_id": "",
    "notes": "",
    "regions": [],
    "url": "https://cloud.baidu.com/"
   },
   "bytedance_doubao": {
    "available": false,
    "fine_tuning": false,
    "gated": false,
    "model_id": "",
    "notes": "",
    "regions": [],
    "url": "https://www.volcengine.com/"
   },
   "cerebras": {
    "available": false,
    "fine_tuning": false,
    "gated": false,
    "model_id": "",
    "notes": "",
    "regions": [],
    "url": "https://www.cerebras.ai/"
   },
   "chatgpt": {
    "available": false,
    "fine_tuning": false,
    "gated": false,
    "model_id": "",
    "notes": "",
    "regions": [],
    "url": "https://chat.openai.com/"
   },
   "claude_ai": {
    "available": false,
    "fine_tuning": false,
    "gated": false,
    "model_id": "",
    "notes": "",
    "regions": [],
    "url": "https://claude.ai/"
   },
   "cohere": {
    "available": false,
    "fine_tuning": false,
    "gated": false,
    "model_id": "",
    "notes": "",
    "regions": [],
    "url": "https://cohere.com/"
   },
   "copilot_microsoft": {
    "available": false,
    "fine_tuning": false,
    "gated": false,
    "model_id": "",
    "notes": "",
    "regions": [],
    "url": "https://copilot.microsoft.com/"
   },
   "cursor": {
    "available": false,
    "fine_tuning": false,
    "gated": false,
    "model_id": "",
    "notes": "",
    "regions": [],
    "url": "https://cursor.com/"
   },
   "deepinfra": {
    "available": false,
    "fine_tuning": false,
    "gated": false,
    "model_id": "",
    "notes": "",
    "regions": [],
    "url": "https://deepinfra.com/"
   },
   "deepseek": {
    "available": false,
    "fine_tuning": false,
    "gated": false,
    "model_id": "",
    "notes": "",
    "regions": [],
    "url": "https://platform.deepseek.com/"
   },
   "fireworks_ai": {
    "available": false,
    "fine_tuning": false,
    "gated": false,
    "model_id": "",
    "notes": "",
    "regions": [],
    "url": "https://fireworks.ai/"
   },
   "gemini_app": {
    "available": false,
    "fine_tuning": false,
    "gated": false,
    "model_id": "",
    "notes": "",
    "regions": [],
    "url": "https://gemini.google.com/"
   },
   "github_copilot": {
    "available": false,
    "fine_tuning": false,
    "gated": false,
    "model_id": "",
    "notes": "",
    "regions": [],
    "url": "https://github.com/features/copilot"
   },
   "google_vertex_ai": {
    "available": false,
    "fine_tuning": false,
    "gated": false,
    "model_id": "",
    "notes": "",
    "regions": [],
    "url": "https://cloud.google.com/vertex-ai"
   },
   "gpt4all": {
    "available": false,
    "fine_tuning": false,
    "gated": false,
    "model_id": "",
    "notes": "",
    "regions": [],
    "url": "https://gpt4all.io/"
   },
   "grok_xai": {
    "available": false,
    "fine_tuning": false,
    "gated": false,
    "model_id": "",
    "notes": "",
    "regions": [],
    "url": "https://x.ai/"
   },
   "groq": {
    "available": false,
    "fine_tuning": false,
    "gated": false,
    "model_id": "",
    "notes": "",
    "regions": [],
    "url": "https://groq.com/"
   },
   "huggingface": {
    "available": false,
    "fine_tuning": false,
    "gated": false,
    "model_id": "",
    "notes": "The official TypeSafe organisation exists and publishes no weights (\"None public yet\", read 2026-09-20). Third-party reproductions are other organisations' artifacts, not TypeSafe's (MODEL-82).",
    "regions": [],
    "url": "https://huggingface.co/typesafe"
   },
   "ibm_watsonx": {
    "available": false,
    "fine_tuning": false,
    "gated": false,
    "model_id": "",
    "notes": "",
    "regions": [],
    "url": "https://www.ibm.com/watsonx"
   },
   "jan_ai": {
    "available": false,
    "fine_tuning": false,
    "gated": false,
    "model_id": "",
    "notes": "",
    "regions": [],
    "url": "https://jan.ai/"
   },
   "kaggle_models": {
    "available": false,
    "fine_tuning": false,
    "gated": false,
    "model_id": "",
    "notes": "",
    "regions": [],
    "url": "https://www.kaggle.com/models"
   },
   "lm_studio": {
    "available": false,
    "fine_tuning": false,
    "gated": false,
    "model_id": "",
    "notes": "",
    "regions": [],
    "url": "https://lmstudio.ai/"
   },
   "meta_ai": {
    "available": false,
    "fine_tuning": false,
    "gated": false,
    "model_id": "",
    "notes": "",
    "regions": [],
    "url": "https://www.meta.ai/"
   },
   "minimax": {
    "available": false,
    "fine_tuning": false,
    "gated": false,
    "model_id": "",
    "notes": "",
    "regions": [],
    "url": "https://www.minimax.chat/"
   },
   "mistral_plateforme": {
    "available": false,
    "fine_tuning": false,
    "gated": false,
    "model_id": "",
    "notes": "",
    "regions": [],
    "url": "https://console.mistral.ai/"
   },
   "mlx_community": {
    "available": false,
    "fine_tuning": false,
    "gated": false,
    "model_id": "",
    "notes": "",
    "regions": [],
    "url": "https://huggingface.co/mlx-community"
   },
   "modelscope": {
    "available": false,
    "fine_tuning": false,
    "gated": false,
    "model_id": "",
    "notes": "",
    "regions": [],
    "url": "https://modelscope.cn/"
   },
   "moonshot_kimi": {
    "available": false,
    "fine_tuning": false,
    "gated": false,
    "model_id": "",
    "notes": "",
    "regions": [],
    "url": "https://www.moonshot.cn/"
   },
   "nvidia_nim": {
    "available": false,
    "fine_tuning": false,
    "gated": false,
    "model_id": "",
    "notes": "",
    "regions": [],
    "url": "https://build.nvidia.com/"
   },
   "ollama": {
    "available": false,
    "fine_tuning": false,
    "gated": false,
    "model_id": "",
    "notes": "",
    "regions": [],
    "url": "https://ollama.com/"
   },
   "open_webui": {
    "available": false,
    "fine_tuning": false,
    "gated": false,
    "model_id": "",
    "notes": "",
    "regions": [],
    "url": "https://openwebui.com/"
   },
   "openrouter": {
    "available": false,
    "fine_tuning": false,
    "gated": false,
    "model_id": "",
    "notes": "",
    "regions": [],
    "url": "https://openrouter.ai/"
   },
   "other_platforms": [
    {
     "available": true,
     "fine_tuning": false,
     "gated": false,
     "model_id": "",
     "notes": "Vercel AI Gateway. Availability, not authorship (MODEL-82). Vercel's own blog of 2026-09-18 reports adoption by its paid teams; the id it is called by on that gateway is not established.",
     "regions": [],
     "url": "https://vercel.com/ai-gateway"
    }
   ],
   "perplexity": {
    "available": false,
    "fine_tuning": false,
    "gated": false,
    "model_id": "",
    "notes": "",
    "regions": [],
    "url": "https://www.perplexity.ai/"
   },
   "poe": {
    "available": false,
    "fine_tuning": false,
    "gated": false,
    "model_id": "",
    "notes": "",
    "regions": [],
    "url": "https://poe.com/"
   },
   "primary_provider": {
    "api_endpoint": "https://api.typesafe.ai/v1/systemone",
    "data_residency": null,
    "data_residency_disclosure": "unresearched",
    "data_residency_source": null,
    "env_vars": [],
    "fedramp_authorized": false,
    "free_tier": false,
    "free_tier_details": "",
    "hipaa_eligible": false,
    "model_id_on_platform": "jev-1.13.0",
    "name": "TypeSafe AI",
    "npm_package": "",
    "platform_url": "https://typesafe.ai/",
    "rate_limit_rpm": 1200,
    "rate_limit_tpm": null,
    "regions": [],
    "sla_uptime": "",
    "soc2_compliant": false
   },
   "qwen_alibaba": {
    "available": false,
    "fine_tuning": false,
    "gated": false,
    "model_id": "",
    "notes": "",
    "regions": [],
    "url": "https://www.alibabacloud.com/en/solutions/generative-ai/qwen"
   },
   "raycast": {
    "available": false,
    "fine_tuning": false,
    "gated": false,
    "model_id": "",
    "notes": "",
    "regions": [],
    "url": "https://www.raycast.com/"
   },
   "replicate": {
    "available": false,
    "fine_tuning": false,
    "gated": false,
    "model_id": "",
    "notes": "",
    "regions": [],
    "url": "https://replicate.com/"
   },
   "sambanova": {
    "available": false,
    "fine_tuning": false,
    "gated": false,
    "model_id": "",
    "notes": "",
    "regions": [],
    "url": "https://sambanova.ai/"
   },
   "samsung_gauss": {
    "available": false,
    "fine_tuning": false,
    "gated": false,
    "model_id": "",
    "notes": "",
    "regions": [],
    "url": "https://www.samsung.com/"
   },
   "snowflake_cortex": {
    "available": false,
    "fine_tuning": false,
    "gated": false,
    "model_id": "",
    "notes": "",
    "regions": [],
    "url": "https://www.snowflake.com/en/data-cloud/cortex/"
   },
   "stability_ai": {
    "available": false,
    "fine_tuning": false,
    "gated": false,
    "model_id": "",
    "notes": "",
    "regions": [],
    "url": "https://stability.ai/"
   },
   "tencent_hunyuan": {
    "available": false,
    "fine_tuning": false,
    "gated": false,
    "model_id": "",
    "notes": "",
    "regions": [],
    "url": "https://cloud.tencent.com/"
   },
   "tii_falcon": {
    "available": false,
    "fine_tuning": false,
    "gated": false,
    "model_id": "",
    "notes": "",
    "regions": [],
    "url": "https://falconllm.tii.ae/"
   },
   "together_ai": {
    "available": false,
    "fine_tuning": false,
    "gated": false,
    "model_id": "",
    "notes": "",
    "regions": [],
    "url": "https://www.together.ai/"
   },
   "upstage_solar": {
    "available": false,
    "fine_tuning": false,
    "gated": false,
    "model_id": "",
    "notes": "",
    "regions": [],
    "url": "https://www.upstage.ai/"
   },
   "zero_one_ai": {
    "available": false,
    "fine_tuning": false,
    "gated": false,
    "model_id": "",
    "notes": "",
    "regions": [],
    "url": "https://www.01.ai/"
   },
   "zhipu_glm": {
    "available": false,
    "fine_tuning": false,
    "gated": false,
    "model_id": "",
    "notes": "",
    "regions": [],
    "url": "https://www.zhipuai.cn/"
   }
  },
  "benchmarks": {
   "benchmark_as_of": "",
   "benchmark_notes": "TypeSafe publishes no benchmark numbers for Jev (docs.typesafe.ai, read 2026-09-20), so this card carries none and the model is returned unranked with insufficient_benchmark_evidence. That is the correct result, not a defect: an empty score set is no evidence, and the generative suites (MMLU, GPQA, SWE-bench) do not apply to a model that generates nothing. Third-party evaluations exist; only their existence was checked, not their results, so none is cited here.",
   "benchmark_source": "",
   "evidence": [],
   "scores": {}
  },
  "capabilities": {
   "agent_capabilities": {
    "autonomous_execution": false,
    "code_execution": false,
    "file_system_access": false,
    "long_running_tasks": false,
    "memory_management": false,
    "self_delegation": false,
    "web_browsing": false
   },
   "coding": {
    "agentic_coding": false,
    "code_completion": false,
    "code_review": false,
    "debugging": false,
    "documentation": false,
    "fill_in_middle": false,
    "languages": [],
    "lsp_integration": false,
    "multi_file_editing": false,
    "overall": null,
    "refactoring": false,
    "repository_understanding": false,
    "test_generation": false
   },
   "creative": {
    "instruction_following": null,
    "storytelling": null,
    "summarization": null,
    "technical_writing": null,
    "writing": null
   },
   "domain_specific": {
    "financial_knowledge": null,
    "legal_knowledge": null,
    "medical_knowledge": null,
    "scientific_knowledge": null
   },
   "language": {
    "long_context_retrieval": null,
    "multilingual": false,
    "num_languages": null,
    "strong_languages": [],
    "translation_quality": null
   },
   "reasoning": {
    "causal": false,
    "chain_of_thought": false,
    "logical": false,
    "mathematical": false,
    "multi_step": false,
    "overall": null,
    "planning": false,
    "scientific": false,
    "self_correction": false,
    "spatial": false,
    "temporal": false,
    "think_budget_control": false
   },
   "safety_alignment": {
    "alignment_approach": "",
    "content_safety_tier": null,
    "guardrail_builtin": false,
    "jailbreak_resistance": null,
    "refusal_rate": null
   },
   "tool_use": {
    "computer_use": false,
    "function_calling": false,
    "mcp_compatible": false,
    "multi_turn_tool_use": false,
    "overall": null,
    "parallel_tool_calls": false,
    "tool_error_recovery": false,
    "tool_selection_accuracy": null
   }
  },
  "card_author": "modelspec",
  "card_created": "2026-09-20",
  "card_schema_version": "3.0",
  "card_updated": "2026-09-20",
  "cost": {
   "batch_input": null,
   "batch_output": null,
   "cache_read": null,
   "cache_write": null,
   "embedding_per_million": null,
   "finetune_hosting_per_hour": null,
   "finetune_per_million_tokens": null,
   "free_tier": false,
   "free_tier_limits": "",
   "input": 0.042,
   "input_audio": null,
   "input_image": null,
   "note": "USD per million input tokens; output tokens are not charged (docs.typesafe.ai/models.md, read 2026-09-20). The schema prices tokens, and a caller of this model budgets per decision: one request carries state plus questions and returns typed answers. That per-decision price is not representable here \u2014 see the card body and MODEL-97.",
   "output": 0.0,
   "output_audio": null,
   "output_image": null,
   "output_video_per_sec": null,
   "reasoning": null,
   "reranking_per_million": null
  },
  "deployment": {
   "api_only": true,
   "custom_hardware": [],
   "fine_tuning_methods": [],
   "fine_tuning_supported": false,
   "hardware_profiles": {
    "dgx_spark_128gb": {
     "best_quant": "",
     "fits": null,
     "inference_engine": "",
     "max_context_at_quant": null,
     "notes": "",
     "prompt_tps": null,
     "ram_usage_gb": null,
     "tokens_per_sec": null,
     "ttft_ms": null,
     "vram_usage_gb": null
    },
    "macbook_air_m4_24gb": {
     "best_quant": "",
     "fits": null,
     "inference_engine": "",
     "max_context_at_quant": null,
     "notes": "",
     "prompt_tps": null,
     "ram_usage_gb": null,
     "tokens_per_sec": null,
     "ttft_ms": null,
     "vram_usage_gb": null
    },
    "macbook_m4_pro_64gb": {
     "best_quant": "",
     "fits": null,
     "inference_engine": "",
     "max_context_at_quant": null,
     "notes": "",
     "prompt_tps": null,
     "ram_usage_gb": null,
     "tokens_per_sec": null,
     "ttft_ms": null,
     "vram_usage_gb": null
    },
    "nvidia_5090_32gb": {
     "best_quant": "",
     "fits": null,
     "inference_engine": "",
     "max_context_at_quant": null,
     "notes": "",
     "prompt_tps": null,
     "ram_usage_gb": null,
     "tokens_per_sec": null,
     "ttft_ms": null,
     "vram_usage_gb": null
    }
   },
   "local_inference": false,
   "quantizations_available": [],
   "runtimes": {
    "core_ml": false,
    "exllamav2": false,
    "gguf": false,
    "llama_cpp": false,
    "lm_studio": false,
    "mlx": false,
    "nim": false,
    "ollama": false,
    "ollama_tag": "",
    "onnx": false,
    "sglang": false,
    "transformers": false,
    "triton": false,
    "trt_llm": false,
    "vllm": false
   },
   "self_hostable": false
  },
  "display_name": "Jev 1.13",
  "downselect": {
   "approval_authority": "",
   "clearance_tags": [],
   "compliance_tags": [],
   "cost_tier": "",
   "custom_notes": "",
   "custom_score": null,
   "defense_tags": [],
   "eval_status": null,
   "next_review_date": "",
   "review_date": "",
   "reviewed_by": "",
   "risk_tier": null,
   "sovereignty_tags": [],
   "use_case_tags": []
  },
  "family": "jev",
  "inference_performance": {
   "api_latency_p50_ms": null,
   "api_latency_p99_ms": null,
   "api_tps_input": null,
   "api_tps_output": null,
   "api_ttft_ms": null,
   "context_speed_degradation": "",
   "generation_time_sec": null,
   "quality_per_dollar": null,
   "quality_per_watt": null
  },
  "last_updated": "",
  "licensing": {
   "academic_use": "unspecified",
   "acceptable_use_policy_url": "",
   "commercial_use": "unspecified",
   "commercial_use_conditions": "",
   "commercial_use_source": null,
   "defense_use": "unspecified",
   "export_control_notes": "",
   "geographic_restrictions": [],
   "government_use": "unspecified",
   "license_type": "proprietary",
   "license_url": "",
   "medical_use": "unspecified",
   "not_for_all_audiences": false,
   "open_weights": false,
   "origin_country": "US",
   "origin_org_type": "private",
   "tos_url": ""
  },
  "lineage": {
   "adapter_rank": null,
   "adapter_type": "",
   "base_model": "",
   "base_model_relation": null,
   "co2_emissions_kg": null,
   "co2_source": "",
   "energy_kwh": null,
   "library_name": "",
   "merge_models": [],
   "training_compute_flops": null,
   "training_cost_estimate": "",
   "training_data_cutoff": "",
   "training_data_tokens": null,
   "training_datasets": [],
   "training_hardware": "",
   "training_method": "other",
   "training_time": ""
  },
  "modalities": {
   "audio": {
    "asr_languages": [],
    "input_supported": false,
    "max_audio_duration_sec": null,
    "music_generation": false,
    "music_understanding": false,
    "output_supported": false,
    "realtime_streaming": false,
    "speaker_diarization": false,
    "tts_languages": [],
    "tts_voices": null,
    "voice_cloning": false
   },
   "document": {
    "form_understanding": false,
    "layout_analysis": false,
    "max_pages": null,
    "pdf_native": false,
    "table_extraction": false
   },
   "embeddings": {
    "batch_size_max": null,
    "dimension_options": [],
    "dimensions": null,
    "dimensions_configurable": false,
    "instruction_aware": false,
    "max_input_tokens": null,
    "normalized": null,
    "similarity_metric": "",
    "supported": false
   },
   "image_generation": {
    "aspect_ratios": [],
    "controlnet_support": false,
    "img2img": false,
    "inpainting": false,
    "lora_support": false,
    "max_resolution": "",
    "outpainting": false,
    "style_control": false,
    "supported": false,
    "text_rendering_quality": ""
   },
   "input": [
    "text"
   ],
   "output": [
    "classifications",
    "scores"
   ],
   "reranking": {
    "cross_encoder": null,
    "max_input_length": null,
    "max_input_pairs": null,
    "supported": false
   },
   "text": {
    "context_window": 64000,
    "fill_in_middle": null,
    "json_mode": null,
    "max_input_tokens": 64000,
    "max_output_tokens": null,
    "streaming": null,
    "system_prompt": false
   },
   "video": {
    "audio_sync": false,
    "input_supported": false,
    "max_fps": null,
    "max_input_duration_sec": null,
    "max_output_duration_sec": null,
    "max_resolution": "",
    "output_supported": false,
    "temporal_reasoning": false
   },
   "vision": {
    "chart_reading": false,
    "handwriting": false,
    "max_image_resolution": "",
    "max_images_per_request": null,
    "object_counting": false,
    "object_detection": false,
    "ocr": false,
    "spatial_reasoning": false,
    "supported": false,
    "video_frames": false,
    "visual_grounding": false
   }
  },
  "model_id": "typesafe/jev-1-13",
  "model_subtypes": [],
  "model_type": "decision-model",
  "pipeline_tag": "",
  "provider": "typesafe",
  "provider_display": "TypeSafe AI",
  "release_date": "",
  "risk_governance": {
   "accountable_and_transparent": "",
   "adversarial_robustness": "untested",
   "bias_evaluation": {
    "conducted": false,
    "known_biases": [],
    "methodology": "",
    "results_summary": ""
   },
   "explainable_and_interpretable": "",
   "fair_with_bias_managed": "",
   "incident_history": [],
   "known_failure_modes": [
    "arithmetic and quantitative comparison",
    "dates and elapsed time",
    "indirection, where the question is about the question",
    "adversarial or manipulative content inside the supplied state",
    "asking it to generate rather than to judge"
   ],
   "privacy": {
    "ccpa_compliant": null,
    "data_processing_location": [],
    "data_retention_policy": "",
    "gdpr_compliant": null,
    "hipaa_eligible": null,
    "pii_handling": "",
    "training_data_pii_scrubbed": null
   },
   "privacy_enhanced": "",
   "regulatory": {
    "eu_ai_act_risk_level": null,
    "fedramp_level": "",
    "iso_42001_certified": null,
    "nist_rmf_profile": "",
    "soc2_type2": null
   },
   "safe": "",
   "secure_and_resilient": "",
   "supply_chain": {
    "ai_bom_available": false,
    "model_provenance_documented": false,
    "reproducible": false,
    "third_party_dependencies": [],
    "training_data_transparency": ""
   },
   "valid_and_reliable": ""
  },
  "sources": {
   "arena_url": "",
   "artificial_analysis_url": "",
   "arxiv_url": "",
   "github_url": "",
   "huggingface_url": "",
   "last_scraped_benchmarks": "2026-09-20",
   "last_scraped_huggingface": "2026-09-20",
   "last_scraped_models_dev": "",
   "last_scraped_pricing": "2026-09-20",
   "models_dev_url": "",
   "ollama_url": "",
   "paper_url": "",
   "provider_docs_url": "https://docs.typesafe.ai/"
  },
  "status": "active",
  "tags": [
   "system-one"
  ],
  "version": "jev-1.13.0"
 }
}