Claire Foulquier-Gazagnes

Optimist who worries • Making AI open @Reflection_ai • Previously @Etalab @Google • Feminist, obviously

[Claire Foulquier-Gazagnes @_cmfg]

Back from a 2 years X diet. Recently joined @reflection_ai and want to keep the pulse on #openweight #AI #inference. What have I missed?

[Reflection @reflection_ai]

One week since we dropped the weights for Beam. 10 trillion tokens served to date.

Try the model for free: https://t.co/ry8O6JPlz6

[Claire Foulquier-Gazagnes @_cmfg]

@dpatil @kevinweil @manohar_paluri Bravo DJ, this is such a worthwhile mission!

[🇺🇦 Alex Polozov @alexpolozov]

Beam is now #1 Western open model on the (truly private) Vals AI eval 🚀 Beating Muse Spark Open, Mistral Large 4, and Inkling.

[Artificial Analysis @ArtificialAnlys]

Reflection’s first model, Beam, scores 30 on the Artificial Analysis Intelligence Index, placing it #2 among US open weights models

Beam is an open weights model with 501B total and 23B active parameters, released on @huggingface

Congratulations to @reflection_ai on the release!

Key takeaways:

➤ Second among US open weights models: Beam scores 30, behind only Meta's Muse Spark Open (39) and ahead of yesterday’s NVIDIA Nemotron 3.5 Super VL (26) release. Outside China, only Muse Spark Open and the UAE's K2 Horizon (31) score higher

➤ Token efficient for its intelligence: Beam uses ~26k output tokens per Intelligence Index task compared to K2 Horizon, one point ahead, which uses twice as many (~52k). Proprietary models still set the frontier, with GPT-6 Astra (max) scoring 53 on the same budget (~27k)

➤ Strong on agentic workflows, weaker in the terminal: Beam scores 48.5% on AutomationBench-AA, 11 percentage points ahead of K2 Horizon (37.2%) and level with Qwen3.8 27B (48.2%). Terminal-Bench 4.0 is a weak point, where Beam scores 3.0%, behind Qwen3.8 27B (5.6%) and Muse Spark Open (8.1%)

➤ Knows more, guesses more: On AA-Omniscience it answers 33% of questions correctly, well ahead of K2 Horizon (18%) and Qwen3.8 27B (16%). But its hallucination rate is 47%, against 26% for K2 Horizon and 30% for Qwen3.8 27B, giving Beam a net index score of 1.5, ahead of both peers (−3.0 and −10.0)

Other model details
Context window: 1M tokens
Modalities: Text
Parameters: 501B / 23B

[Claire Foulquier-Gazagnes @_cmfg]

Try out #Beam: https://t.co/7NqiM3YYmg

[Reflection @reflection_ai]

Beam is now available across platforms, with open weights under the Apache 2.0 license - in collaboration with an ecosystem of 40+ partners.

We're releasing:
- Weights, tokenizer, and reference implementation
- BF16, FP8 and NVFP4 weights and a DFlash2 drafter
- The open source harness we built and trained Beam on: Mirror CLI

Weights and model card on Hugging Face: https://t.co/tPqWpk9RXL

[Reflection @reflection_ai]

Beam is now available across platforms, with open weights under the Apache 2.0 license - in collaboration with an ecosystem of 40+ partners.

We're releasing:
- Weights, tokenizer, and reference implementation
- BF16, FP8 and NVFP4 weights and a DFlash2 drafter
- The open source harness we built and trained Beam on: Mirror CLI

Weights and model card on Hugging Face: https://t.co/tPqWpk9RXL

[Reflection @reflection_ai]

Introducing Beam: a highly efficient agentic open model with 501B total parameters and 23B active.

- Frontier reasoning efficiency
- Advances the Western open frontier on coding & agentic tasks
- Trained end-to-end from scratch

Full weights release this month.

Learn more about Beam: https://t.co/c3Qx2cpM8G

[Claire Foulquier-Gazagnes @_cmfg]

@MonopSav Merci mais ça ne marche toujours pas après désinstallation et réinstallation

[Laure Pressac 💙💛 @LaurePressac]

oui on vit dans un monde fou, où je vous invite a contribuer à une cagnotte lancée par @Nabilla
parce que là vraiment, cette femme est notre héroïne
https://t.co/8VBv6EEfDp

[Claire Foulquier-Gazagnes @_cmfg]

@RaphaelleRL @LaurePressac Tout notre business plan s'effondre !

[Claire Foulquier-Gazagnes @_cmfg]

@alik41 @br_ttany I won't deny that the bureaucracy is frustrating and I had my fair share of struggles.
The flip side is that you get much more support from the gov for your young kids than most countries (high daycare subsidy, free schooling, social aid if you qualify, health benefits...).

[Claire Foulquier-Gazagnes @_cmfg]

@br_ttany 3 downsides of Paris imo:
1) To truly integrate, you will need to learn French, easy for kids maybe a bit less for you
2) Subway has a lot of stairs and not many restaurants have high chairs / changing tables
3) Central Park equivalents are further away, on the edge of Paris

[Claire Foulquier-Gazagnes @_cmfg]

@br_ttany I have lived in NYC (without kids though) and I feel safer in Paris (firearms are extremely regulated here). On the cleanliness, it is the same as NYC, big city life, some rats (bigger in NYC) and dog poos.

[Claire Foulquier-Gazagnes @_cmfg]

@br_ttany Strongly recommend moving there before September so you get them both enrolled for the "rentrée". It is way harder to get a daycare spot if you don't start in Sept.
The big downside of Paris is the not stroller friendly subway with lots of stairs but if your 2 yo walks well..

[Claire Foulquier-Gazagnes @_cmfg]

@br_ttany Paris is a great option for affordability of childcare and diversity of activities.
Your 2 yo could be enrolled in a daycare, max 950€ / month or so for up to 10 hours a day. Your 4 yo would be enrolled in a primary school, which is free, except for the canteen.

[Fred Mazéas @FredMazeas]

Ça y est. C'est... fini. #Paris2024

[Matilde Meslin @MatildeMeslin]

C’est l’heure du best of de la Phryge (ou Clitolympique, pour les intimes), l’icône de ces Jeux Olympiques @Paris2024 ⬇️

[Alister @SchnockRevue]

Sept minutes parfaites...
#ceremoniedouverture