AI Index / LLM application development tools / History

LLM application development tools — AI Recommendation History

What AI assistants recommended for llm application development tools, date by date. Every ranking we publish is frozen the day it publishes, so this is a record of what was said — not a re-computation of what today's data would say about the past.

Record: 26 dated snapshots · Aug 14, 2026 – Sep 8, 2026

Answer archive Methodology Data studies JSON

What changed

Computed by comparing each dated snapshot with the one before it. Share moves smaller than 1pp are treated as noise and left out — the same threshold that suppresses trend arrows on the ranking.

Sep 8, 2026vs Sep 7, 2026

No change above the noise floor.

Sep 7, 2026vs Sep 6, 2026

No change above the noise floor.

Sep 6, 2026vs Sep 5, 2026

No change above the noise floor.

Sep 5, 2026vs Sep 4, 2026

No change above the noise floor.

Sep 4, 2026vs Sep 3, 2026

No change above the noise floor.

Sep 3, 2026vs Sep 2, 2026

Methodology change: extraction rules 4-7 → 7 between these dates. Differences below may reflect the rule change rather than a change in what AI recommends — see the methodology changelog.

▲ INAzure entered the top 10 (#16 → #5)
▼ OUTLangSmith left the top 10 (#6 → #12)
▲Azure moved #16 → #5
▼LangSmith moved #6 → #12
▲Pinecone moved #10 → #7
▼Langfuse moved #7 → #9
▼Semantic Kernel moved #8 → #10
▼Microsoft moved #5 → #6
▲Ollama moved #9 → #8
▼LangSmith -15.3pp (36.7% → 21.4%)
▲Azure +13.7pp (18.4% → 32.1%)
▼Haystack -11.2pp (46.9% → 35.7%)
▼LangChain -8.2pp (79.6% → 71.4%)
▼Microsoft -8.1pp (36.7% → 28.6%)
▼Langfuse -7.7pp (32.7% → 25%)
▼Semantic Kernel -7.2pp (28.6% → 21.4%)
▼LlamaIndex -4.1pp (75.5% → 71.4%)
▲Pinecone +4.1pp (24.5% → 28.6%)
▲LangGraph +3.1pp (46.9% → 50%)
▼Ollama -1.5pp (26.5% → 25%)

Sep 2, 2026vs Sep 1, 2026

No change above the noise floor.

Sep 1, 2026vs Aug 31, 2026

No change above the noise floor.

Aug 31, 2026vs Aug 30, 2026

▲Microsoft moved #6 → #5
▼LangSmith moved #5 → #6
▲Ollama moved #10 → #9
▼Pinecone moved #9 → #10
▲Microsoft +2.8pp (33.9% → 36.7%)
▼LangChain -2.5pp (82.1% → 79.6%)
▲Haystack +2.3pp (44.6% → 46.9%)
▲Langfuse +2.3pp (30.4% → 32.7%)
▼Pinecone -2.3pp (26.8% → 24.5%)
▲Semantic Kernel +1.8pp (26.8% → 28.6%)
▼LangGraph -1.3pp (48.2% → 46.9%)

Aug 30, 2026vs Aug 29, 2026

▲Semantic Kernel moved #9 → #8
▼Pinecone moved #8 → #9
▼LangSmith -3.8pp (41.3% → 37.5%)
▲Ollama +3pp (23.8% → 26.8%)
▼Pinecone -1.8pp (28.6% → 26.8%)
▲Haystack +1.7pp (42.9% → 44.6%)
▼Langfuse -1.3pp (31.7% → 30.4%)
▼LlamaIndex -1.2pp (76.2% → 75%)
▼Microsoft -1pp (34.9% → 33.9%)

Aug 29, 2026vs Aug 28, 2026

No change above the noise floor.

Aug 28, 2026vs Aug 27, 2026

No change above the noise floor.

Aug 27, 2026vs Aug 26, 2026

▲ INOllama entered the top 10 (#12 → #10)
▼ OUTQdrant left the top 10 (#10 → #11)
▲Ollama moved #12 → #10
▲Haystack moved #5 → #4
▼LangSmith moved #4 → #5
▲Pinecone moved #9 → #8
▼Semantic Kernel moved #8 → #9
▼Qdrant moved #10 → #11
▼LangSmith -3.3pp (44.6% → 41.3%)
▲LangGraph +3pp (44.6% → 47.6%)
▼Langfuse -2.2pp (33.9% → 31.7%)
▼Semantic Kernel -1.6pp (28.6% → 27%)
▼LangChain -1.4pp (83.9% → 82.5%)
▼Qdrant -1.2pp (25% → 23.8%)

Aug 26, 2026vs Aug 25, 2026

Methodology change: extraction rules 4 → 4-7 between these dates. Differences below may reflect the rule change rather than a change in what AI recommends — see the methodology changelog.

▲ INQdrant entered the top 10 (#14 → #10)
▼ OUTDify left the top 10 (#10 → #18)
▼Dify moved #10 → #18
▲Qdrant moved #14 → #10
▲LangGraph moved #5 → #3
▲Microsoft moved #8 → #6
▼LangSmith moved #3 → #4
▼Haystack moved #4 → #5
▼Langfuse moved #6 → #7
▼Semantic Kernel moved #7 → #8
▲Microsoft +7.1pp (28.6% → 35.7%)
▼Dify -6.9pp (26.5% → 19.6%)
▼LangSmith -6.4pp (51% → 44.6%)
▼Haystack -6.1pp (49% → 42.9%)
▼Semantic Kernel -6.1pp (34.7% → 28.6%)
▼LlamaIndex -2.8pp (79.6% → 76.8%)
▼Langfuse -2.8pp (36.7% → 33.9%)
▼LangChain -1.8pp (85.7% → 83.9%)
▲LangGraph +1.7pp (42.9% → 44.6%)

Aug 25, 2026vs Aug 24, 2026

No change above the noise floor.

Aug 24, 2026vs Aug 23, 2026

No change above the noise floor.

Aug 23, 2026vs Aug 22, 2026

No change above the noise floor.

Aug 22, 2026vs Aug 21, 2026

▲ INMicrosoft entered the top 10 (#14 → #8)
▼ OUTvLLM left the top 10 (#10 → #12)
▲Microsoft moved #14 → #8
▼Dify moved #8 → #10
▼vLLM moved #10 → #12
▲LangSmith moved #4 → #3
▼Haystack moved #3 → #4
▲LangSmith +8.1pp (42.9% → 51%)
▲Microsoft +7.2pp (21.4% → 28.6%)
▲Langfuse +3.8pp (32.9% → 36.7%)
▲Semantic Kernel +3.3pp (31.4% → 34.7%)
▲LangGraph +2.9pp (40% → 42.9%)
▲LangChain +2.8pp (82.9% → 85.7%)
▲LlamaIndex +2.5pp (77.1% → 79.6%)
▼Haystack -2.4pp (51.4% → 49%)
▲Pinecone +1.5pp (27.1% → 28.6%)

Aug 21, 2026vs Aug 20, 2026

No change above the noise floor.

Aug 20, 2026vs Aug 19, 2026

No change above the noise floor.

Aug 19, 2026vs Aug 18, 2026

No change above the noise floor.

Aug 18, 2026vs Aug 17, 2026

No change above the noise floor.

Aug 17, 2026vs Aug 16, 2026

No change above the noise floor.

Aug 16, 2026vs Aug 15, 2026

No change above the noise floor.

Aug 15, 2026vs Aug 14, 2026

No change above the noise floor.

The ranking, date by date

Top 10 on each archived date, exactly as published. Cite a date and it stays true: snapshots are append-only, and a correction appends a new revision rather than overwriting the original.

Sep 8, 202628 sampled answers · window 2026-08-11 – 2026-09-08 · rules v7

#1LangChain71.4%
#2LlamaIndex71.4%
#3LangGraph50%
#4Haystack35.7%
#5Azure32.1%
#6Microsoft28.6%
#7Pinecone28.6%
#8Ollama25%
#9Langfuse25%
#10Semantic Kernel21.4%

Sep 7, 202628 sampled answers · window 2026-08-10 – 2026-09-07 · rules v7

#1LangChain71.4%
#2LlamaIndex71.4%
#3LangGraph50%
#4Haystack35.7%
#5Azure32.1%
#6Microsoft28.6%
#7Pinecone28.6%
#8Ollama25%
#9Langfuse25%
#10Semantic Kernel21.4%

Sep 6, 202628 sampled answers · window 2026-08-09 – 2026-09-06 · rules v7

#1LangChain71.4%
#2LlamaIndex71.4%
#3LangGraph50%
#4Haystack35.7%
#5Azure32.1%
#6Microsoft28.6%
#7Pinecone28.6%
#8Ollama25%
#9Langfuse25%
#10Semantic Kernel21.4%

Sep 5, 202628 sampled answers · window 2026-08-08 – 2026-09-05 · rules v7

#1LangChain71.4%
#2LlamaIndex71.4%
#3LangGraph50%
#4Haystack35.7%
#5Azure32.1%
#6Microsoft28.6%
#7Pinecone28.6%
#8Ollama25%
#9Langfuse25%
#10Semantic Kernel21.4%

Sep 4, 202628 sampled answers · window 2026-08-07 – 2026-09-04 · rules v7

#1LangChain71.4%
#2LlamaIndex71.4%
#3LangGraph50%
#4Haystack35.7%
#5Azure32.1%
#6Microsoft28.6%
#7Pinecone28.6%
#8Ollama25%
#9Langfuse25%
#10Semantic Kernel21.4%

Sep 3, 202628 sampled answers · window 2026-08-06 – 2026-09-03 · rules v7

#1LangChain71.4%
#2LlamaIndex71.4%
#3LangGraph50%
#4Haystack35.7%
#5Azure32.1%
#6Microsoft28.6%
#7Pinecone28.6%
#8Ollama25%
#9Langfuse25%
#10Semantic Kernel21.4%

Sep 2, 202649 sampled answers · window 2026-08-05 – 2026-09-02 · rules v4-7

#1LangChain79.6%
#2LlamaIndex75.5%
#3LangGraph46.9%
#4Haystack46.9%
#5Microsoft36.7%
#6LangSmith36.7%
#7Langfuse32.7%
#8Semantic Kernel28.6%
#9Ollama26.5%
#10Pinecone24.5%

Sep 1, 202649 sampled answers · window 2026-08-04 – 2026-09-01 · rules v4-7

#1LangChain79.6%
#2LlamaIndex75.5%
#3LangGraph46.9%
#4Haystack46.9%
#5Microsoft36.7%
#6LangSmith36.7%
#7Langfuse32.7%
#8Semantic Kernel28.6%
#9Ollama26.5%
#10Pinecone24.5%

Aug 31, 202649 sampled answers · window 2026-08-03 – 2026-08-31 · rules v4-7

#1LangChain79.6%
#2LlamaIndex75.5%
#3LangGraph46.9%
#4Haystack46.9%
#5Microsoft36.7%
#6LangSmith36.7%
#7Langfuse32.7%
#8Semantic Kernel28.6%
#9Ollama26.5%
#10Pinecone24.5%

Aug 30, 202656 sampled answers · window 2026-08-02 – 2026-08-30 · rules v4-7

#1LangChain82.1%
#2LlamaIndex75%
#3LangGraph48.2%
#4Haystack44.6%
#5LangSmith37.5%
#6Microsoft33.9%
#7Langfuse30.4%
#8Semantic Kernel26.8%
#9Pinecone26.8%
#10Ollama26.8%

Aug 29, 202663 sampled answers · window 2026-08-01 – 2026-08-29 · rules v4-7

#1LangChain82.5%
#2LlamaIndex76.2%
#3LangGraph47.6%
#4Haystack42.9%
#5LangSmith41.3%
#6Microsoft34.9%
#7Langfuse31.7%
#8Pinecone28.6%
#9Semantic Kernel27%
#10Ollama23.8%

Aug 28, 202663 sampled answers · window 2026-07-31 – 2026-08-28 · rules v4-7

#1LangChain82.5%
#2LlamaIndex76.2%
#3LangGraph47.6%
#4Haystack42.9%
#5LangSmith41.3%
#6Microsoft34.9%
#7Langfuse31.7%
#8Pinecone28.6%
#9Semantic Kernel27%
#10Ollama23.8%

Aug 27, 202663 sampled answers · window 2026-07-30 – 2026-08-27 · rules v4-7

#1LangChain82.5%
#2LlamaIndex76.2%
#3LangGraph47.6%
#4Haystack42.9%
#5LangSmith41.3%
#6Microsoft34.9%
#7Langfuse31.7%
#8Pinecone28.6%
#9Semantic Kernel27%
#10Ollama23.8%

Aug 26, 202656 sampled answers · window 2026-07-29 – 2026-08-26 · rules v4-7

#1LangChain83.9%
#2LlamaIndex76.8%
#3LangGraph44.6%
#4LangSmith44.6%
#5Haystack42.9%
#6Microsoft35.7%
#7Langfuse33.9%
#8Semantic Kernel28.6%
#9Pinecone28.6%
#10Qdrant25%

Aug 25, 202649 sampled answers · window 2026-07-28 – 2026-08-25 · rules v4

#1LangChain85.7%
#2LlamaIndex79.6%
#3LangSmith51%
#4Haystack49%
#5LangGraph42.9%
#6Langfuse36.7%
#7Semantic Kernel34.7%
#8Microsoft28.6%
#9Pinecone28.6%
#10Dify26.5%

Aug 24, 202649 sampled answers · window 2026-07-27 – 2026-08-24 · rules v4

#1LangChain85.7%
#2LlamaIndex79.6%
#3LangSmith51%
#4Haystack49%
#5LangGraph42.9%
#6Langfuse36.7%
#7Semantic Kernel34.7%
#8Microsoft28.6%
#9Pinecone28.6%
#10Dify26.5%

Aug 23, 202649 sampled answers · window 2026-07-26 – 2026-08-23 · rules v4

#1LangChain85.7%
#2LlamaIndex79.6%
#3LangSmith51%
#4Haystack49%
#5LangGraph42.9%
#6Langfuse36.7%
#7Semantic Kernel34.7%
#8Microsoft28.6%
#9Pinecone28.6%
#10Dify26.5%

Aug 22, 202649 sampled answers · window 2026-07-25 – 2026-08-22 · rules v4

#1LangChain85.7%
#2LlamaIndex79.6%
#3LangSmith51%
#4Haystack49%
#5LangGraph42.9%
#6Langfuse36.7%
#7Semantic Kernel34.7%
#8Microsoft28.6%
#9Pinecone28.6%
#10Dify26.5%

Aug 21, 202670 sampled answers · window 2026-07-24 – 2026-08-21 · rules v4

#1LangChain82.9%
#2LlamaIndex77.1%
#3Haystack51.4%
#4LangSmith42.9%
#5LangGraph40%
#6Langfuse32.9%
#7Semantic Kernel31.4%
#8Dify27.1%
#9Pinecone27.1%
#10vLLM24.3%

Aug 20, 202670 sampled answers · window 2026-07-23 – 2026-08-20 · rules v4

#1LangChain82.9%
#2LlamaIndex77.1%
#3Haystack51.4%
#4LangSmith42.9%
#5LangGraph40%
#6Langfuse32.9%
#7Semantic Kernel31.4%
#8Dify27.1%
#9Pinecone27.1%
#10vLLM24.3%

Aug 19, 202670 sampled answers · window 2026-07-22 – 2026-08-19 · rules v4

#1LangChain82.9%
#2LlamaIndex77.1%
#3Haystack51.4%
#4LangSmith42.9%
#5LangGraph40%
#6Langfuse32.9%
#7Semantic Kernel31.4%
#8Dify27.1%
#9Pinecone27.1%
#10vLLM24.3%

Aug 18, 202670 sampled answers · window 2026-07-21 – 2026-08-18 · rules v4

#1LangChain82.9%
#2LlamaIndex77.1%
#3Haystack51.4%
#4LangSmith42.9%
#5LangGraph40%
#6Langfuse32.9%
#7Semantic Kernel31.4%
#8Dify27.1%
#9Pinecone27.1%
#10vLLM24.3%

Aug 17, 202670 sampled answers · window 2026-07-20 – 2026-08-17 · rules v4

#1LangChain82.9%
#2LlamaIndex77.1%
#3Haystack51.4%
#4LangSmith42.9%
#5LangGraph40%
#6Langfuse32.9%
#7Semantic Kernel31.4%
#8Dify27.1%
#9Pinecone27.1%
#10vLLM24.3%

Aug 16, 202670 sampled answers · window 2026-07-19 – 2026-08-16 · rules v4

#1LangChain82.9%
#2LlamaIndex77.1%
#3Haystack51.4%
#4LangSmith42.9%
#5LangGraph40%
#6Langfuse32.9%
#7Semantic Kernel31.4%
#8Dify27.1%
#9Pinecone27.1%
#10vLLM24.3%

Aug 15, 202670 sampled answers · window 2026-07-18 – 2026-08-15 · rules v4

#1LangChain82.9%
#2LlamaIndex77.1%
#3Haystack51.4%
#4LangSmith42.9%
#5LangGraph40%
#6Langfuse32.9%
#7Semantic Kernel31.4%
#8Dify27.1%
#9Pinecone27.1%
#10vLLM24.3%

Aug 14, 202670 sampled answers · window 2026-07-17 – 2026-08-14 · rules v4

#1LangChain82.9%
#2LlamaIndex77.1%
#3Haystack51.4%
#4LangSmith42.9%
#5LangGraph40%
#6Langfuse32.9%
#7Semantic Kernel31.4%
#8Dify27.1%
#9Pinecone27.1%
#10vLLM24.3%

Read any date directly

Every archived date is machine-readable at /api/index/llm-application-development-tools/as-of/<YYYY-MM-DD>, which returns the ranking as it stood, with the extraction version and revision it was published under. Dates before the record begins return no_record rather than a recomputation — we do not reconstruct history on demand and present it as what we said. See what the engines actually said.

[ORBATOR]