Skip to content

Tested prompt · RAG and answering from documents

A fact from one section: every AI model's reply, tested

We sent this everyday RAG and answering from documents prompt to all 16 models in llmwise, the same way the app sends a message, and checked every reply the same way. Here's each one as it came, with whether it passed, what it cost and how long it took.

Based on 16 of our test runs on , through OpenRouter with the app's own prompt and settings. Updated .

Short answer

All 16 models passed this RAG and answering from documents prompt's check (key facts). The cheapest reply that passed was GPT-6 Luna's, at $0.000074; the fastest, GLM 5.3 Flash's in 0.7 s. The dearest reply, Claude Fable 5.1's, cost 181 times as much ($0.0134).

The prompt, as sent, and its check

Checked by key facts, the same way for every model.

A fact from one section (everyday)

[9 lines every RAG and answering from documents prompt of ours shares, word for word: the whole prompt, on the methods page]
Question: How many weeks of parental leave do employees get, and at what pay?

The answer must state “16 weeks”, “full pay”, “§4”.

Exactly what this prompt's replies are checked against, with every other prompt of our test runs.

Every model's result

All 16 models on this prompt, in catalog order.

Every model's reply to “A fact from one section”
ModelResultCostTimeReply
Claude Fable 5.1AnthropicPassed: Stated all 3 facts with the section.$0.01343.5 s45 tokens
Claude Opus 5.5AnthropicPassed: Stated all 3 facts with the section.$0.00542.5 s48 tokens
Claude Sonnet 5.5AnthropicPassed: Stated all 3 facts with the section.$0.00271.4 s46 tokens
Claude Sonnet 5AnthropicPassed: Stated all 3 facts with the section.$0.00191.7 s15 tokens
Claude Haiku 4.5AnthropicPassed: Stated all 3 facts with the section.$0.000911.1 s40 tokens
GPT-6 AstraOpenAIPassed: Stated all 3 facts with the section.$0.00731.3 s18 tokens
GPT-6 SolOpenAIPassed: Stated all 3 facts with the section.$0.00151.1 s18 tokens
GPT-6 LunaOpenAIPassed: Stated all 3 facts with the section.$0.0000741.3 s19 tokens
Gemini 3.1 Pro (preview)GooglePassed: Stated all 3 facts with the section.$0.00455.3 s15 tokens
Gemini 3.8 FlashGooglePassed: Stated all 3 facts with the section.$0.000583.1 s14 tokens
DeepSeek V4.1 FlashDeepSeekPassed: Stated all 3 facts with the section.$0.000150.9 s9 tokens
DeepSeek V4 ProDeepSeekPassed: Stated all 3 facts with the section.$0.000242.2 s15 tokens
Grok 4.7xAIPassed: Stated all 3 facts with the section.$0.00252.7 s14 tokens
Kimi K3MoonshotPassed: Stated all 3 facts with the section.$0.00106.5 s42 tokens
GLM 5.3Z.aiPassed: Stated all 3 facts with the section.$0.000260.9 s11 tokens
GLM 5.3 FlashZ.aiPassed: Stated all 3 facts with the section.$0.000200.7 s17 tokens

Cost: what OpenRouter charged us for the reply. Time: from sending to the whole reply. Reply: its length in tokens, thinking not counted. In llmwise you pay per message, not per token: each of these replies counted as one message on Pro.

Every reply

Every reply passed: here they are from the cheapest up.

  1. GPT-6 Luna

    Passed: Stated all 3 facts with the section.

    Employees get 16 weeks of parental leave at full pay. (§4)

    642 tokens in, 19 out · 1.3 s · $0.0001 · 1 message on Pro · answered by openai/gpt-6-luna via OpenAI ·

  2. DeepSeek V4.1 Flash

    Passed: Stated all 3 facts with the section.

    16 weeks at full pay (§4).

    689 tokens in, 22 out (13 of them reasoning) · 0.9 s · $0.0002 · 1 message on Pro · answered by deepseek/deepseek-v4.1-flash via Novita ·

  3. GLM 5.3 Flash

    Passed: Stated all 3 facts with the section.

    Employees get **16 weeks of parental leave at full pay** (§4).

    662 tokens in, 31 out (14 of them reasoning) · 0.7 s · $0.0002 · 1 message on Pro · answered by z-ai/glm-5.3-flash via Wafer ·

  4. DeepSeek V4 Pro

    Passed: Stated all 3 facts with the section.

    Employees get 16 weeks of parental leave at full pay (§4).

    661 tokens in, 80 out (65 of them reasoning) · 2.2 s · $0.0002 · 1 message on Pro · answered by deepseek/deepseek-v4-pro-0813 via StreamLake ·

  5. GLM 5.3

    Passed: Stated all 3 facts with the section.

    16 weeks at full pay (§4).

    661 tokens in, 11 out · 0.9 s · $0.0003 · 1 message on Pro · answered by z-ai/glm-5.3 via Baidu ·

  6. Gemini 3.8 Flash

    Passed: Stated all 3 facts with the section.

    Employees get 16 weeks of leave at full pay (§4).

    701 tokens in, 14 out · 3.1 s · $0.0006 · 1 message on Pro · answered by google/gemini-3.8-flash via Google ·

  7. Claude Haiku 4.5

    Passed: Stated all 3 facts with the section.

    According to §4, every parent gets **16 weeks of leave at full pay**. The leave can be split into two blocks, both taken within the child's first year.

    713 tokens in, 40 out · 1.1 s · $0.0009 · 1 message on Pro · answered by anthropic/claude-haiku-4.5 via Amazon Bedrock ·

  8. Kimi K3

    Passed: Stated all 3 facts with the section.

    Employees get 16 weeks of parental leave at full pay. It can be split into two blocks, both taken within the child's first year. (§4)

    735 tokens in, 51 out (9 of them reasoning) · 6.5 s · $0.0010 · 1 message on Pro · answered by moonshotai/kimi-k3 via Morph ·

  9. GPT-6 Sol

    Passed: Stated all 3 facts with the section.

    Employees get 16 weeks of parental leave at full pay (§4).

    642 tokens in, 18 out · 1.1 s · $0.0015 · 1 message on Pro · answered by openai/gpt-6-sol via OpenAI ·

  10. Claude Sonnet 5

    Passed: Stated all 3 facts with the section.

    16 weeks at full pay (§4).

    898 tokens in, 15 out · 1.7 s · $0.0019 · 1 message on Pro · answered by anthropic/claude-sonnet-5 via Claude Platform on AWS ·

  11. Grok 4.7

    Passed: Stated all 3 facts with the section.

    Employees get 16 weeks of parental leave at full pay (§4).

    1,876 tokens in, 174 out (160 of them reasoning) · 2.7 s · $0.0025 · 1 message on Pro · answered by x-ai/grok-4.7 via xAI ·

  12. Claude Sonnet 5.5

    Passed: Stated all 3 facts with the section.

    Employees get 16 weeks of parental leave at full pay. It can be split into two blocks, both taken within the child's first year. (§4)

    902 tokens in, 46 out · 1.4 s · $0.0027 · 1 message on Pro · answered by anthropic/claude-sonnet-5.5 via Anthropic ·

  13. Gemini 3.1 Pro

    Passed: Stated all 3 facts with the section.

    Employees get 16 weeks of parental leave at full pay (§4).

    701 tokens in, 261 out (246 of them reasoning) · 5.3 s · $0.0045 · 1 message on Pro · answered by google/gemini-3.1-pro-preview via Google ·

  14. Claude Opus 5.5

    Passed: Stated all 3 facts with the section.

    Employees get 16 weeks of parental leave at full pay. It can be split into two blocks, and both must be taken within the child's first year (§4).

    900 tokens in, 48 out · 2.5 s · $0.0054 · 1 message on Pro · answered by anthropic/claude-opus-5.5 via Claude Platform on AWS ·

  15. GPT-6 Astra

    Passed: Stated all 3 facts with the section.

    Employees get 16 weeks of parental leave at full pay (§4).

    642 tokens in, 18 out · 1.3 s · $0.0073 · 1 message on Pro · answered by openai/gpt-6-astra via OpenAI ·

  16. Claude Fable 5.1

    Passed: Stated all 3 facts with the section.

    Every parent gets 16 weeks of leave at full pay. It can be split into two blocks, both taken within the child's first year (§4).

    900 tokens in, 45 out · 3.5 s · $0.0134 · 1 message on Pro · answered by anthropic/claude-fable-5.1 via Anthropic ·

More RAG and answering from documents prompts

The other RAG and answering from documents prompts, each with every model's reply, and the results across all five.

Questions

Which AI does best on “A fact from one section”?

All 16 models passed this RAG and answering from documents prompt's check (key facts). The cheapest reply that passed was GPT-6 Luna's, at $0.000074; the fastest, GLM 5.3 Flash's in 0.7 s. The dearest reply, Claude Fable 5.1's, cost 181 times as much ($0.0134).

What does a reply to “A fact from one section” cost?

Through the models' APIs, what OpenRouter charged us ran from $0.000074 (GPT-6 Luna) to $0.0134 (Claude Fable 5.1) for this prompt. In llmwise you don't pay by the token: a reply like these counts as one message on Pro, whichever model answers.

Claude, GPT, Gemini, DeepSeek, Grok, Kimi, and GLM, in one chat.

See what a message costs before you send it. Free is 5 messages to try; sign in with an email link, no password or card.