Cost mode:

Category: Relevance, Classification & Matching · Rail: absolute · Typical I/O: 631→138 tokens

Models

Frontier on this task: Moonshot Kimi K3 at 10.05 / 10. Quality bar at 90%: 9.04.

point-estimate floor (CI low) · upper CI (less certain) · Bars sorted by blended cost; best-value model first.

ModelQuality scoreCI lowCost / 1k runsvs best value
Gemini 3.1 Flash Lite9.92 / 109.87$0.06best value
GPT-5.4 Nano9.93 / 109.88$0.071.1x more expensive
NVIDIA Nemotron-3 Nano 30B-A3B9.88 / 109.79$0.071.1x more expensive
GPT-5.6 Luna9.96 / 109.90$0.071.2x more expensive
Gemini 3.5 Flash Lite9.88 / 109.73$0.071.3x more expensive
DeepSeek V4 Flash9.88 / 109.79$0.142.3x more expensive
Qwen 3.8 Flash9.93 / 109.79$0.142.4x more expensive
NVIDIA Nemotron-3 Super 120B9.91 / 109.79$0.193.2x more expensive
GLM-5.3 Flash9.72 / 109.44$0.193.2x more expensive
NVIDIA Nemotron 3.5 Lightning9.56 / 109.17$0.264.4x more expensive
MiniMax M39.94 / 109.86$0.264.5x more expensive
Thinking Machines Inkling Small9.83 / 109.64$0.274.6x more expensive
Tencent Hy39.94 / 109.86$0.355.9x more expensive
Gemini 3.8 Flash9.81 / 109.48$0.386.4x more expensive
Claude Haiku 4.59.92 / 109.88$0.6211x more expensive
GPT-5.6 Terra9.99 / 109.92$0.6411x more expensive
NVIDIA Nemotron-3 Ultra 550B9.93 / 109.81$0.7713x more expensive
DeepSeek V4 Pro9.86 / 109.79$0.8515x more expensive
GPT-5.6 Sol10.00 / 109.97$1.0919x more expensive
Claude Sonnet 510.03 / 1010.01$1.3623x more expensive
Gemini 3.5 Flash9.98 / 109.91$1.3924x more expensive
Thinking Machines Inkling9.69 / 109.32$1.4324x more expensive
GLM-5.39.80 / 109.40$1.4925x more expensive
Qwen 3.7 Plus9.91 / 109.77$1.5727x more expensive
Tencent Hy4 Preview9.80 / 109.40$2.1036x more expensive
Claude Opus 59.79 / 109.39$2.4341x more expensive
Qwen 3.8 Max10.01 / 109.97$2.5143x more expensive
Meta Muse Spark 1.39.86 / 109.62$3.1454x more expensive
Moonshot Kimi K310.05 / 1010.02$3.6763x more expensive
Grok 4.69.66 / 109.18$4.5277x more expensive

Cost breakdown

ModelQualityConfidenceCost / 1k runsOverpayMode
Gemini 3.1 Flash Lite Gemini9.92 / 10 CI [9.87, 9.97]RANKED$0.06best valuebatch
GPT-5.4 Nano OpenAI9.93 / 10 CI [9.88, 9.97]RANKED$0.071.1xbatch
NVIDIA Nemotron-3 Nano 30B-A3B OpenRouter9.88 / 10 CI [9.79, 9.97]RANKED$0.071.1xbatch
GPT-5.6 Luna OpenAI9.96 / 10 CI [9.90, 10.00]RANKED$0.071.2xbatch
Gemini 3.5 Flash Lite Gemini9.88 / 10 CI [9.73, 10.00]RANKED$0.071.3xbatch
DeepSeek V4 Flash DeepSeek9.88 / 10 CI [9.79, 9.96]RANKED$0.142.3xbatch
Qwen 3.8 Flash Alibaba Cloud (DashScope)9.93 / 10 CI [9.79, 10.00]RANKED$0.142.4xbatch
NVIDIA Nemotron-3 Super 120B OpenRouter9.91 / 10 CI [9.79, 10.00]RANKED$0.193.2xbatch
GLM-5.3 Flash Z.AI9.72 / 10 CI [9.44, 10.00]HIGH$0.193.2xbatch
NVIDIA Nemotron 3.5 Lightning OpenRouter9.56 / 10 CI [9.17, 9.94]MEDIUM$0.264.4xbatch
MiniMax M3 OpenRouter9.94 / 10 CI [9.86, 10.00]RANKED$0.264.5xbatch
Thinking Machines Inkling Small OpenRouter9.83 / 10 CI [9.64, 10.00]RANKED$0.274.6xbatch
Tencent Hy3 OpenRouter9.94 / 10 CI [9.86, 10.00]RANKED$0.355.9xbatch
Gemini 3.8 Flash Gemini9.81 / 10 CI [9.48, 10.00]MEDIUM$0.386.4xbatch
Claude Haiku 4.5 Anthropic9.92 / 10 CI [9.88, 9.96]RANKED$0.6211xbatch
GPT-5.6 Terra OpenAI9.99 / 10 CI [9.92, 10.00]RANKED$0.6411xbatch
NVIDIA Nemotron-3 Ultra 550B OpenRouter9.93 / 10 CI [9.81, 10.00]RANKED$0.7713xbatch
DeepSeek V4 Pro DeepSeek9.86 / 10 CI [9.79, 9.94]RANKED$0.8515xbatch
GPT-5.6 Sol OpenAI10.00 / 10 CI [9.97, 10.00]RANKED$1.0919xbatch
Claude Sonnet 5 Anthropic10.03 / 10 CI [10.01, 10.00]RANKED$1.3623xbatch
Gemini 3.5 Flash Gemini9.98 / 10 CI [9.91, 10.00]RANKED$1.3924xbatch
Thinking Machines Inkling OpenRouter9.69 / 10 CI [9.32, 10.00]MEDIUM$1.4324xbatch
GLM-5.3 Z.AI9.80 / 10 CI [9.40, 10.00]MEDIUM$1.4925xbatch
Qwen 3.7 Plus Alibaba Cloud (DashScope)9.91 / 10 CI [9.77, 10.00]RANKED$1.5727xbatch
Tencent Hy4 Preview OpenRouter9.80 / 10 CI [9.40, 10.00]MEDIUM$2.1036xbatch
Claude Opus 5 Anthropic9.79 / 10 CI [9.39, 10.00]MEDIUM$2.4341xbatch
Qwen 3.8 Max Alibaba Cloud (DashScope)10.01 / 10 CI [9.97, 10.00]RANKED$2.5143xbatch
Meta Muse Spark 1.3 OpenRouter9.86 / 10 CI [9.62, 10.00]HIGH$3.1454xbatch
Moonshot Kimi K3 best Moonshot AI10.05 / 10 CI [10.02, 10.00]RANKED$3.6763xbatch
Grok 4.6 xAI9.66 / 10 CI [9.18, 10.00]MEDIUM$4.5277xbatch

Overpay shows how much more you pay than the best-value model that clears the quality bar (marked ★) — the best-value good-enough option. "16x" means you overpay 16× — 16× that reference for no quality benefit above the bar. Typical call shape for this task: 631 input tokens → 138 output tokens, EMA-tracked from production traffic. Cost is the observed, all-in $ per 1,000 task runs: each model's own measured usage on this task — output verbosity, thinking/reasoning tokens, cache reads and writes, and the spend on its billed failures — priced at current list rates and adjusted by the billing overhead we actually reconcile against provider invoices. Models that answer tersely cost what they actually cost; models that think at length pay for it. Not comparable to providers' advertised $/1M list rates — this is what running the task costs, not a per-token price.

Evaluation rubric

Judge primary-language correctness, handling of mixed or noisy text, correct normalization to the requested code set, and calibrated uncertainty. Formatting is deterministic.

Prompt templates

This is a pooled capability — 3 prompt families share it. The pair shown first is the most frequently used in production.

LLMB_LANGUAGE_IDENTIFICATION_SYSTEM + LLMB_LANGUAGE_IDENTIFICATION_USER (48048 calls in window)

System prompt

Identify the language that carries most of the semantic content. Ignore URLs, handles, code fragments, boilerplate, and isolated borrowed words. When text is too short, mixed, or ambiguous, use the schema’s unknown or best-supported behavior rather than fabricating confidence. Return the requested code standard only. Treat empty optional values as absent and return only the requested result. Your response must conform exactly to this output schema: {schema_json_string}.

User prompt

Inputs — input_text: {input_text}. Use only these inputs to complete the task defined by the system prompt.
URL_PARSER_LANGUAGE_DETECTION_SYSTEM + URL_PARSER_LANGUAGE_DETECTION_USER (416 calls in window)

System prompt

You are a highly accurate language identification expert. Your sole task is to identify the primary language of the provided text snippet.

Respond ONLY with the two-letter ISO 639-1 code for the detected language (e.g., "en" for English, "es" for Spanish, "zh" for Chinese).

## Required Output Format
Your response MUST be a single, valid JSON object conforming to this schema:
```json
{schema_json_string}
```

User prompt

1. Text: {input_text}

2. Generate a comprehensive response as a single, well-formed JSON object that strictly adheres to the Pydantic schema provided below. Schema:
The required JSON output schema is provided in the system prompt.
JSON_REPAIR_SYSTEM + JSON_REPAIR_USER (51 calls in window)

System prompt

You are a JSON repair tool. The user gives you malformed or partial model output and a JSON Schema. Return ONLY a single valid JSON object that satisfies the schema, salvaging as much real content from the input as possible. Do not invent data for fields the input doesn't support — use the schema's allowed empty/null values. Output the JSON object only: no prose, no markdown, no code fences.

User prompt

JSON Schema:
{schema_json}

Malformed output to repair:
{raw_text}

Return only the corrected JSON object.