Hardcore Reviews
Hardcore Reviews

Tencent's 770B Flagship Activates Only 49B: Five Open-Weight Flagships Compared, and Total Parameters Don't Decide Deployment Cost

Hy4 preview (770B/49B) pushes the open-source flagship parameter race to a new high, but total parameters don't decide deployment cost: active parameters save compute, while weight residency consumes VRAM. This review lines up five open-weight flagships - Hy4 preview, GLM-5.3, Kimi K3 (2.8T), DeepSeek V4 (1.6T reported) and Qwen3.8-Max (2.4T) - across active/total ratio, context, license, VRAM threshold (engineering estimates) and API price snapshots. Division of labor with the Aug 27 price review: that one ran the API math at the 320B tier, this one runs the parameter and deployment-threshold math at 700B-2.8T. Five scenario verdicts: pick Hy4 for the newest (Apache 2.0 + MTP speculative decoding + FP8-friendly), K3 for raw scale, GLM/DeepSeek for mature ecosystems, Qwen for Alibaba-compliance stacks, and for everyone: check per-token cost and sparse attention before total parameters.

Published August 29, 20269 min read
<!-- open-source-flagship-llm-comparison-review | review | Tencent's 770B Flagship Activates Only 49B: Five Open-Weight Flagships Compared, and Total Parameters Don't Decide Deployment Cost -->

On August 28, Tencent Hunyuan released its new flagship Hy4 preview along with its weights: 770B total parameters, only 49B activated per token. Sitting at the same table are Kimi K3 at 2.8 trillion, DeepSeek V4-Pro at roughly 1.6 trillion, Qwen3.8-Max at 2.4T, and the post-training-route GLM-5.3. Total parameters span 770B to 2.8T -- a 3.6x spread -- but when you open the deployment ledger, the variable that actually decides cost is not total size. It is the one most coverage skips: the activation ratio and sparse attention.

Scope first: this is a representative comparison, not hands-on benchmarking. Specs, architecture, and eval figures come from official releases and public reporting; prices and repo stats are a 2026-08-29 snapshot, with the sourcing basis labeled per item (official release / media report / converted / engineering estimate / vendor internal blind eval). Defer to official repos and pricing pages. Not investment advice. One division of labor to note: our lite flagship price comparison covers the API unit-price and cache math of 320B-class lite flagships; this article does not redo the API price war. It covers the parameter, architecture, and deployment-threshold ledger of 700B-2.8T open-weight flagships. Two articles, two ledgers -- read both before you choose.

1. The Parameter Baseline: Five-Way Spec Table

DimensionTencent Hy4 previewKimi K3 (Moonshot)DeepSeek V4-ProQwen3.8-Max (Alibaba)Zhipu GLM-5.3
Total parameters770B (official release)2.8T (official release)~1.6T (our earlier reporting)2.4T (our earlier fact-check)Not collected
Active params/token49B (official release)Not collectedNot collectedNot collectedNot collected
Context window1M (official release)1M (official release)1M (our earlier reporting)1M (our earlier fact-check)Not collected
Attention / architecture highlightsGated DSA sparse attention + IndexCache; 78 layers (1 dense FFN + 77 MoE; 256 routed + 1 shared expert, top-8 + shared)KDA hybrid linear attention + attention residualsNot collectedNot collectedPost-training scaling (base unchanged)
Speculative decodingNative MTP layer (official release)Not collectedNot collectedNot collectedNot collected
Weight format / quantizationBF16 + FP8 dual format (official release)MXFP4 quantized weights (official release)Not collectedNot collectedNot collected
LicenseApache 2.0 (official release)Commercial use allowed (terms per official repo)Not collectedPer official repo (not verified in this snapshot)Open weight (repo/license per official channels)
Weight channelsHugging Face / ModelScope / GitCode / CNBHugging Face (weights opened 2026-07-27)Not collectedNot collectedZhipu official repo (THUDM)

Three first-glance takeaways:

  1. Active parameters are buried in the press release. Only Hy4 preview puts its active parameter count (49B) front and center; for the other four, active counts are either not collected or inconsistently reported. That is no accident: the activation ratio (active/total) is the lifeline of inference cost. Hy4's 49B/770B is about 6.4% -- each token runs roughly 1/16 of the network. Total parameters decide how much the model knows; active parameters decide how much compute each token burns. Press releases love the former; the ledger cares about the latter.
  2. Sparse attention has become flagship standard equipment. Hy4's Gated DSA + IndexCache and K3's KDA hybrid linear attention attack the same problem: attention compute and KV cache blow-up at 1M context. The previous generation paid double inference cost to run 1M windows; this generation compresses that curve with sparse/linear hybrids. When a spec sheet says "supports 1M context," the real cost behind that sentence depends entirely on the attention architecture.
  3. Licenses and quantization formats are diverging. Hy4 ships Apache 2.0 with BF16/FP8 dual weights; K3 ships MXFP4 quantized weights; the GLM line is open but license terms defer to official channels. Quantization format is not a footnote: it directly determines how much VRAM the resident weights occupy (next section does that math).

2. The Deployment Threshold: Activation Saves Compute, Total Size Eats VRAM

The most common misreading of this cohort: small active parameters do not mean small VRAM needs. Our GLM-5.3-Flash integration SOP ran this math: activating 18B saves per-token compute (throughput, latency, power), but the full 320B of weights must still sit in VRAM. At 770B-2.8T scale, the same rule becomes a cluster-scale problem:

ModelResident weight VRAM (converted / engineering estimate)Deployment shapeKey variable
Tencent Hy4 preview~770GB at FP8 / ~1.5TB at BF16 (converted at 1/2 byte per param)Multi-node cluster (official vLLM image starts at TP8)Active 49B saves compute; total 770B sets VRAM; MTP speculative decoding boosts throughput
Kimi K3~1.4TB at MXFP4 (converted at 4 bits per param)Multi-node cluster (earlier reporting: H100-class cluster floor)Total 2.8T sets the VRAM ceiling; KDA compresses KV cache
DeepSeek V4-ProNot collectedNot collectedNot collected
Qwen3.8-MaxNot collectedNot collectedNot collected
Zhipu GLM-5.3Not collectedNot collectedNot collected

Note: every VRAM figure above is a converted estimate (bit-width x parameter count; excludes KV cache and runtime overhead, so real requirements run higher). Defer to each official repo's deployment docs.

  1. The real 770B threshold: a cluster, not a single 8-GPU box. Hy4 preview's FP8 weights run about 770GB; 8x80GB gives 640GB of VRAM, which does not fit. You need at least 10-12 80GB cards or 16 48GB cards (engineering estimate). The official path is the prebuilt vLLM image (vllm/vllm-openai:hy4-preview, starting at TP8) and an SGLang image, with README deployment docs written for cluster scale. For individual developers and small teams, self-hosting stops here.
  2. The payoff of low activation is throughput, not VRAM. 49B active means small per-token compute, and paired with the native MTP speculative decoding layer (official basis), throughput per unit of hardware clearly beats a dense-activation model of the same total size. For offline batch workloads, a low activation ratio saves time and power; for low-latency online serving, it shapes first-token latency. The rule: activation ratio saves compute, total size sets VRAM -- keep the two ledgers separate.
  3. Sparse attention decides the KV cache curve. At 1M context, dense attention's KV cache explodes linearly with sequence length; Hy4's Gated DSA + IndexCache and K3's KDA hybrid linear attention both compress that curve. When choosing, do not stop at "supports 1M" -- look at the attention architecture, which decides your real VRAM and cost curve in long-context scenarios.
  4. The shared threshold of this cohort: nobody self-hosts any of the five on a single card. 770B-2.8T total parameters make self-hosting an institutional game. The realistic play for small teams is API plus open weights as fallback: run on API daily, migrate to self-hosting only for extreme price hikes or compliance scenarios. Migratability is exactly what open weights buy you -- but remember the migration target itself is a cluster-scale threshold.

3. Capability and Price Snapshot: One Blind Eval, One Aggregator Price, Three Caveats

DimensionTencent Hy4 previewKimi K3DeepSeek V4-ProQwen3.8-MaxZhipu GLM-5.3
Representative capability dataVendor internal blind eval 2.99/4.00 (163 experts, 203 engineering tasks, official-release basis)Frontend Code Arena 1679, topping Claude Fable 5 (media citation)AA index 53 (media citation)Not collectedOpen-source first on Terminal Bench 3.0 and others (official release)
API priceOpenRouter snapshot: $0.834/M input, $2.501/M output, $0.042/M cache read (third-party aggregator snapshot 2026-08-28)Output 100 RMB/M (official pricing via media, 3x+ the previous generation)Peak output 27 RMB/M (media report 2026-08-27)Not collectedAPI pricing per Zhipu official channels
Cache price$0.042/M cache read (OpenRouter snapshot)Not collectedV4-Flash off-peak 0.05 RMB/M (media report)Cache hit 1 RMB vs miss 12 RMB (our earlier fact-check)Not collected

Three sourcing caveats:

  1. The 2.99 score is a vendor internal blind eval, not third-party benchmarking. It edges GLM-5.3's 2.92 (46.8% win / 12.8% tie / 40.4% loss) and Kimi K3's 2.94 (51.2% win / 7.9% tie / 40.9% loss), over 163 experts and 203 engineering tasks -- a decent sample, but the evaluator is the vendor itself, which picks the tasks, the experts, and the scoring rubric. Treat it as a reference point, not a verdict; run your own eval set before adopting.
  2. The OpenRouter snapshot price is an aggregator price, not Tencent's official rate card. $0.834/$2.501 per million tokens is a third-party aggregator snapshot from 2026-08-28; aggregators add margins, quotas, and FX conversion. It is not the official pricing page. Verify official rates through Tencent Cloud's official channels -- we did not collect unverified official pricing.
  3. Cache pricing is fragmented; do not compare cache rates directly across vendors. The five use wildly different cache bases: Hy4 cache read $0.042/M (OpenRouter snapshot), DeepSeek V4-Flash off-peak 0.05 RMB/M (media report), Qwen3.8-Max 1 RMB hit vs 12 RMB miss (our earlier fact-check) -- different currencies, billing units, and hit rules. In the Agent era, cache hit rate steers the bill (our lite flagship comparison showed selection conclusions can flip once hit rates exceed 90%). Before switching models, redo your cache math on a unified basis.

4. Pick by Scenario: Five Verdicts

  1. Want the newest open flagship and its architecture dividends: Tencent Hy4 preview. Apache 2.0 license, BF16/FP8 dual weights, native MTP speculative decoding, complete official vLLM/SGLang images -- the open ecosystem can absorb it. The 49B activation throughput dividend suits high-concurrency workloads. For the hands-on path, see our Hy4 preview self-hosting SOP (same batch).
  2. Want maximum scale and capability ceiling: Kimi K3. 2.8T total (the first open-source 3-trillion-class model), native vision, 1M context, Frontend Code Arena 1679 at the top. Output at 100 RMB/M is a capability premium tier -- not a value pick.
  3. Want maturity and ecosystem balance: Zhipu GLM-5.3 and DeepSeek V4-Pro. GLM-5.3 took open-source first on Terminal Bench 3.0 and more via post-training scaling, plus emergent security capability (2,436 vulnerabilities found jointly with security teams). V4-Pro leans on 1M context plus Anthropic API compatibility for cheap migration. Both are mature dual-line players: commercial API plus open weights.
  4. Alibaba ecosystem and compliance: Qwen3.8-Max. 1M context, cache hit at 1 RMB vs 12 RMB miss (our earlier fact-check), on top of Alibaba Cloud's ecosystem and enterprise compliance stack. Heavy-cache scenarios make that pricing interesting -- but its open-weight status and broader pricing were not verified in this snapshot; defer to official channels before purchasing.
  5. For every scenario: in the 770B-2.8T era, the activation ratio and sparse attention are the deciding variables of deployment cost. Total parameters set resident VRAM (self-hosting threshold), active parameters set per-token compute (throughput/latency), attention architecture sets the long-context cost curve -- keep three ledgers, choose by actual per-token cost, and stop ranking models by total size.

FAQ

Q1: 770B vs 2.8 trillion is a 3.6x gap -- does bigger total size always win? A1: No. Total parameters set resident VRAM (self-hosting threshold), active parameters set per-token compute (throughput/latency/power), and sparse attention sets the long-context cost curve. Treat them separately. Hy4 preview's 770B activates only 49B (about 6.4%) -- its activation ratio is actually the advantage. Choosing by total size alone gets you "buying big and running small." Put all three variables into your own workload and compute actual per-token cost.

Q2: Can Hy4 preview's 2.99 score be taken as "open-source No. 1"? A2: Not directly. The 2.99/4.00 comes from Tencent's internal blind eval (163 experts, 203 engineering tasks); the vendor picks tasks, experts, and the rubric. GLM-5.3's 2.92 and Kimi K3's 2.94 are from the same internal eval. A 0.02-0.07 gap is within sample noise -- use it as a coordinate, not a conclusion, and run your own eval set first.

Q3: Can small teams self-host these flagships? A3: Essentially no. 770B at FP8 is about 770GB of weights (converted estimate) and 2.8T at MXFP4 about 1.4TB -- both far beyond a single 8-GPU box's 640GB of VRAM, before KV cache and runtime overhead. Self-hosting is an institutional game. The realistic path for small teams: API daily, open weights as a migration fallback for extreme hikes or compliance scenarios. Do not put self-hosting on the near-term roadmap.

Q4: Are $0.834/$2.501 on OpenRouter Hy4's official prices? A4: No. That is a third-party aggregator snapshot from 2026-08-28, including platform margin, quotas, and FX conversion -- not Tencent's official rate card. Tencent's official pricing was not verified in this snapshot and is therefore marked not collected; verify via Tencent Cloud's official channels.

Q5: Is anyone on this list worth locking a price with now? A5: Do not lock with anyone. Price data in this snapshot is scattered across bases (aggregator prices, media citations, not-collected entries), the industry is in a pricing-power contest, and rate cards change fast. The right posture: budget by actual per-token cost plus cache hit rate, keep multiple API channels plus open-weight fallback in your architecture, and renegotiate quarterly.


References

This is a representative comparison compiled from public reporting and official releases, snapshot 2026-08-29 (OpenRouter as of 2026-08-28), not hands-on benchmarking and not investment advice. Specs, licenses, and prices may change at any time; verify against official repos and pricing pages before purchasing.

This article is AI-assisted and human-edited. Last updated: 2026-08-29

FAQ

770B vs 2.8 trillion is a 3.6x gap -- does bigger total size always win?
No. Total parameters set resident VRAM (self-hosting threshold), active parameters set per-token compute (throughput/latency/power), and sparse attention sets the long-context cost curve. Treat them separately. Hy4 preview's 770B activates only 49B (about 6.4%) -- its activation ratio is actually the advantage. Choosing by total size alone gets you "buying big and running small." Put all three variables into your own workload and compute actual per-token cost.
Can Hy4 preview's 2.99 score be taken as "open-source No. 1"?
Not directly. The 2.99/4.00 comes from Tencent's internal blind eval (163 experts, 203 engineering tasks); the vendor picks tasks, experts, and the rubric. GLM-5.3's 2.92 and Kimi K3's 2.94 are from the same internal eval. A 0.02-0.07 gap is within sample noise -- use it as a coordinate, not a conclusion, and run your own eval set first.
Can small teams self-host these flagships?
Essentially no. 770B at FP8 is about 770GB of weights (converted estimate) and 2.8T at MXFP4 about 1.4TB -- both far beyond a single 8-GPU box's 640GB of VRAM, before KV cache and runtime overhead. Self-hosting is an institutional game. The realistic path for small teams: API daily, open weights as a migration fallback for extreme hikes or compliance scenarios. Do not put self-hosting on the near-term roadmap.
Are $0.834/$2.501 on OpenRouter Hy4's official prices?
No. That is a third-party aggregator snapshot from 2026-08-28, including platform margin, quotas, and FX conversion -- not Tencent's official rate card. Tencent's official pricing was not verified in this snapshot and is therefore marked not collected; verify via Tencent Cloud's official channels.
Is anyone on this list worth locking a price with now?
Do not lock with anyone. Price data in this snapshot is scattered across bases (aggregator prices, media citations, not-collected entries), the industry is in a pricing-power contest, and rate cards change fast. The right posture: budget by actual per-token cost plus cache hit rate, keep multiple API channels plus open-weight fallback in your architecture, and renegotiate quarterly.

Related

Hardcore Reviews

Nine Days After DeepSeek's 350% Price Hike, Zhipu Cuts In at Half Price: A Five-Way Lite-Flagship API Cost Comparison - and Why Coders Shouldn't Switch Just Yet

On August 17, DeepSeek raised V4-Flash peak output pricing 350% (2 to 9 RMB per million tokens) and introduced peak/off-peak pricing; nine days later Zhipu answered with GLM-5.3-Flash (2.8 RMB output). This comparison runs the numbers by scenario: for zero-cache chat/RAG/long-document workloads GLM is nearly half price (27%-31% during the promo), but for coding agents with 90%+ cache-hit rates, DeepSeek V4-Flash off-peak is about 32% cheaper (estimated basis) - the vendor that raised prices may still be the cheaper one for its heaviest users. During peak hours (9-12, 14-18) GLM's flat pricing wins biggest; Kimi K3 at 100 RMB/M output is a capability premium, not value; and the industry ledger shows price hikes are collective (DeepSeek net loss 715M RMB, Zhipu +83% pricing with +400% volume). Five scenario-based verdicts included.

Aug 27, 20269 min read