NLP Lineage

8 April 2026 · Reasoning

Muse Spark

Muse Spark

Meta Superintelligence Labs

Date
8 April 2026
What the date marks
Model release
Kind
Model
Organisation
Meta
Access
closed, hosted
Other dates
v1.1 2026-07-09
Threads
In the guided tour
Chapter 8. The weights get out
Learning phases
3. API usage and prompting

Too recent for its influence to be judged: no later milestone in the data descends from it yet. How levels are decided

Figures quoted here are reported by the organisation that built it, not independently verified.

Read the primary source (ai.meta.com)

Full summary in progress. Until then, this page shows the milestone's record from the research data.

What it changed

Before it

Meta had anchored the open-weight ecosystem since LLaMA in 2023.

The problem

Open releases gave away capability while the returns accrued to competitors and forks.

What it introduced

A closed multimodal reasoning model from a reorganised research group, with no open weights.

Why it matters

The largest Western backer of open weights changed sides, leaving the open frontier largely to Chinese labs and OpenAI's gpt-oss.

What modern models inherit

An open frontier with a very different geography from 2023.

What came next

An open ecosystem centred on DeepSeek, Qwen, Kimi and Zhipu.

Before and after

The milestones it descends from, and the milestones that descend from it. Work developed alongside it is not counted. Weak links are left out unless you include them.

Before: where it came from

Muse Spark has 51 documented ancestors across 9 chapters, reaching back to Recurrent networks in 1990. Showing its direct parents and the 21 turning points among them.

Muse Spark has 51 documented ancestors across 9 chapters, reaching back to Recurrent networks in 1990. Showing all of them.

Muse Spark has 57 recorded ancestors, including indirect ones, across 9 chapters, reaching back to Recurrent networks in 1990. Showing its direct parents and the 21 turning points among them.

Muse Spark has 57 recorded ancestors, including indirect ones, across 9 chapters, reaching back to Recurrent networks in 1990. Showing all of them.

  1. 2026Muse SparkThis milestone.
  2. 2025Llama 4Muse Spark succeeded thisChapter 8. The weights get out
  3. 2024Llama 3.1 405BLlama 4 succeeded thisChapter 8. The weights get out
  4. 2024Llama 3Llama 3.1 405B built directly on thisChapter 8. The weights get out
  5. 2023Llama 2 Turning pointLlama 3 built directly on thisChapter 8. The weights get out
  6. 2023Position interpolationLlama 3.1 405B was influenced by thisChapter 10. Reading more than you remember
  7. 2023DPOLlama 3.1 405B built directly on thisChapter 6. Teaching models what people want
  8. 2023Grouped-query attention Turning pointLlama 3.1 405B built directly on thisChapter 9. Making it run anywhere
  9. 2023LLaMA Turning pointPosition interpolation built directly on thisChapter 8. The weights get out
  10. 2022Chinchilla Turning pointLlama 3.1 405B was influenced by thisChapter 5. Scale becomes a science
  11. 2022InstructGPT Turning pointDPO built directly on thisChapter 6. Teaching models what people want
  12. 2021GopherChinchilla built directly on thisChapter 5. Scale becomes a science
  13. 2021FLANInstructGPT was influenced by thisChapter 6. Teaching models what people want
  14. 2021RoPE Turning pointPosition interpolation built directly on thisChapter 10. Reading more than you remember
  15. 2021Switch TransformerLlama 4 was influenced by thisChapter 5. Scale becomes a science
  16. 2020RLHF for summarisation Turning pointDPO built directly on thisChapter 6. Teaching models what people want
  17. 2020GShard Turning pointSwitch Transformer built directly on thisChapter 5. Scale becomes a science
  18. 2020GPT-3 Turning pointInstructGPT built directly on thisChapter 5. Scale becomes a science
  19. 2020Scaling laws Turning pointChinchilla built directly on thisChapter 5. Scale becomes a science
  20. 2019Multi-query attentionGrouped-query attention built directly on thisChapter 9. Making it run anywhere
  21. 2019T5 Turning pointSwitch Transformer built directly on thisChapter 4. Pretrain once, use everywhere
  22. 2019C4 Turning pointT5 was influenced by thisChapter 4. Pretrain once, use everywhere
  23. 2019RLHF for LMsRLHF for summarisation built directly on thisChapter 6. Teaching models what people want
  24. 2019Sparse Transformer Turning pointGPT-3 built directly on thisChapter 4. Pretrain once, use everywhere
  25. 2019GPT-2 Turning pointT5 was influenced by thisChapter 4. Pretrain once, use everywhere
  26. 2019WebTextGPT-2 was influenced by thisChapter 4. Pretrain once, use everywhere
  27. 2019Transformer-XLRoPE was influenced by thisChapter 4. Pretrain once, use everywhere
  28. 2018BERT Turning pointT5 was influenced by thisChapter 4. Pretrain once, use everywhere
  29. 2018SentencePieceT5 built directly on thisChapter 4. Pretrain once, use everywhere
  30. 2018GPT-1BERT was influenced by this
    GPT-2 built directly on this    Chapter 4. Pretrain once, use everywhere
  31. 2018ELMoBERT was influenced by thisChapter 4. Pretrain once, use everywhere
  32. 2018ULMFiTBERT was influenced by thisChapter 4. Pretrain once, use everywhere
  33. 2017CoVeELMo was influenced by thisChapter 3. Attention, then only attention
  34. 2017Transformer Turning pointT5 built directly on thisChapter 3. Attention, then only attention
  35. 2017RLHF originRLHF for summarisation built directly on thisChapter 6. Teaching models what people want
  36. 2017Sparse MoE Turning pointSwitch Transformer built directly on thisChapter 3. Attention, then only attention
  37. 2016Layer normTransformer built directly on this
    GPT-2 built directly on this    Chapter 3. Attention, then only attention
  38. 2015Residual connectionsTransformer was influenced by thisChapter 3. Attention, then only attention
  39. 2015LM pretraining Turning pointBERT was influenced by thisChapter 2. Words become vectors
  40. 2015BPE tokenizationGPT-2 built directly on this
    LLaMA built directly on this    Chapter 2. Words become vectors
  41. 2015Luong attentionTransformer was influenced by thisChapter 3. Attention, then only attention
  42. 2014AdamTransformer was influenced by thisChapter 2. Words become vectors
  43. 2014GloVeELMo built directly on thisChapter 2. Words become vectors
  44. 2014Seq2seq Turning pointTransformer built directly on thisChapter 2. Words become vectors
  45. 2014Attention Turning pointTransformer built directly on thisChapter 3. Attention, then only attention
  46. 2014GRU / encoder-decoderAttention built directly on thisChapter 2. Words become vectors
  47. 2013word2vecELMo was influenced by thisChapter 2. Words become vectors
  48. 2011NLP from scratchELMo was influenced by thisChapter 1. Language as probability
  49. 2003Neural LMNLP from scratch was influenced by this
    word2vec was influenced by this    Chapter 1. Language as probability
  50. 1997LSTM Turning pointSparse MoE built directly on thisChapter 1. Language as probability
  51. 1991Mixture of expertsSparse MoE built directly on thisChapter 1. Language as probability
  52. 1990Recurrent networksLSTM was influenced by thisChapter 1. Language as probability
  1. 2026Muse SparkThis milestone.
  2. 2025Llama 4Muse Spark succeeded thisChapter 8. The weights get out
  3. 2024Llama 3.1 405BLlama 4 succeeded thisChapter 8. The weights get out
  4. 2024Llama 3Llama 3.1 405B built directly on thisChapter 8. The weights get out
  5. 2023Mixtral 8x7BLlama 4 was influenced by this (indirect influence)Chapter 8. The weights get out
  6. 2023Mistral 7BMixtral 8x7B built directly on thisChapter 8. The weights get out
  7. 2023Llama 2 Turning pointLlama 3 built directly on thisChapter 8. The weights get out
  8. 2023Position interpolationLlama 3.1 405B was influenced by thisChapter 10. Reading more than you remember
  9. 2023DPOLlama 3.1 405B built directly on thisChapter 6. Teaching models what people want
  10. 2023Grouped-query attention Turning pointLlama 3.1 405B built directly on thisChapter 9. Making it run anywhere
  11. 2023LLaMA Turning pointPosition interpolation built directly on thisChapter 8. The weights get out
  12. 2022FlashAttentionMistral 7B was influenced by thisChapter 9. Making it run anywhere
  13. 2022Chinchilla Turning pointLlama 3.1 405B was influenced by thisChapter 5. Scale becomes a science
  14. 2022InstructGPT Turning pointDPO built directly on thisChapter 6. Teaching models what people want
  15. 2021GopherChinchilla built directly on thisChapter 5. Scale becomes a science
  16. 2021FLANInstructGPT was influenced by thisChapter 6. Teaching models what people want
  17. 2021RoPE Turning pointPosition interpolation built directly on thisChapter 10. Reading more than you remember
  18. 2021Switch TransformerLlama 4 was influenced by thisChapter 5. Scale becomes a science
  19. 2020RLHF for summarisation Turning pointDPO built directly on thisChapter 6. Teaching models what people want
  20. 2020BigBirdFlashAttention succeeded thisChapter 10. Reading more than you remember
  21. 2020GShard Turning pointSwitch Transformer built directly on this
    Mixtral 8x7B built directly on this    Chapter 5. Scale becomes a science
  22. 2020GPT-3 Turning pointInstructGPT built directly on thisChapter 5. Scale becomes a science
  23. 2020LongformerMistral 7B was influenced by thisChapter 10. Reading more than you remember
  24. 2020Scaling laws Turning pointChinchilla built directly on thisChapter 5. Scale becomes a science
  25. 2019Multi-query attentionGrouped-query attention built directly on thisChapter 9. Making it run anywhere
  26. 2019T5 Turning pointSwitch Transformer built directly on thisChapter 4. Pretrain once, use everywhere
  27. 2019C4 Turning pointT5 was influenced by thisChapter 4. Pretrain once, use everywhere
  28. 2019RLHF for LMsRLHF for summarisation built directly on thisChapter 6. Teaching models what people want
  29. 2019Sparse Transformer Turning pointMistral 7B was influenced by thisChapter 4. Pretrain once, use everywhere
  30. 2019GPT-2 Turning pointT5 was influenced by thisChapter 4. Pretrain once, use everywhere
  31. 2019WebTextGPT-2 was influenced by this
    C4 was influenced by this (indirect influence)    Chapter 4. Pretrain once, use everywhere
  32. 2019Transformer-XLSparse Transformer was influenced by this (indirect influence)
    RoPE was influenced by this    Chapter 4. Pretrain once, use everywhere
  33. 2018BERT Turning pointT5 was influenced by thisChapter 4. Pretrain once, use everywhere
  34. 2018SentencePieceT5 built directly on thisChapter 4. Pretrain once, use everywhere
  35. 2018GPT-1BERT was influenced by this
    GPT-2 built directly on this    Chapter 4. Pretrain once, use everywhere
  36. 2018ELMoBERT was influenced by thisChapter 4. Pretrain once, use everywhere
  37. 2018ULMFiTBERT was influenced by thisChapter 4. Pretrain once, use everywhere
  38. 2017CoVeELMo was influenced by thisChapter 3. Attention, then only attention
  39. 2017Transformer Turning pointT5 built directly on thisChapter 3. Attention, then only attention
  40. 2017RLHF originRLHF for summarisation built directly on thisChapter 6. Teaching models what people want
  41. 2017ConvS2STransformer was influenced by this (indirect influence)Chapter 3. Attention, then only attention
  42. 2017Sparse MoE Turning pointSwitch Transformer built directly on this
    Mixtral 8x7B was influenced by this    Chapter 3. Attention, then only attention
  43. 2016Layer normTransformer built directly on this
    GPT-2 built directly on this    Chapter 3. Attention, then only attention
  44. 2015Residual connectionsTransformer was influenced by thisChapter 3. Attention, then only attention
  45. 2015LM pretraining Turning pointBERT was influenced by thisChapter 2. Words become vectors
  46. 2015BPE tokenizationGPT-2 built directly on this
    LLaMA built directly on this    Chapter 2. Words become vectors
  47. 2015Luong attentionTransformer was influenced by thisChapter 3. Attention, then only attention
  48. 2014AdamTransformer was influenced by thisChapter 2. Words become vectors
  49. 2014GloVeELMo built directly on thisChapter 2. Words become vectors
  50. 2014Seq2seq Turning pointTransformer built directly on thisChapter 2. Words become vectors
  51. 2014Attention Turning pointTransformer built directly on thisChapter 3. Attention, then only attention
  52. 2014GRU / encoder-decoderAttention built directly on thisChapter 2. Words become vectors
  53. 2013word2vecELMo was influenced by thisChapter 2. Words become vectors
  54. 2011NLP from scratchELMo was influenced by thisChapter 1. Language as probability
  55. 2003Neural LMNLP from scratch was influenced by this
    word2vec was influenced by this    Chapter 1. Language as probability
  56. 1997LSTM Turning pointSparse MoE built directly on thisChapter 1. Language as probability
  57. 1991Mixture of expertsSparse MoE built directly on thisChapter 1. Language as probability
  58. 1990Recurrent networksLSTM was influenced by thisChapter 1. Language as probability

After: what it led to

No documented descendants are recorded for Muse Spark without weak links.

No documented descendants are recorded for Muse Spark without weak links.

No recorded descendants are recorded for Muse Spark.

No recorded descendants are recorded for Muse Spark.

Direct links and their sources

Where it came from

Succeeded

What it led to

None recorded in the sources used.